Flood forecasting method and device based on historical cases, storage medium and computer

Abstract

The application discloses a flood forecasting method and device based on historical cases, a storage medium and a computer. The method comprises the following steps: obtaining flood forecasting data of a to-be-predicted river basin, wherein the flood forecasting data comprises river basin environment data; obtaining a plurality of river basin flood cases, wherein the river basin flood cases comprise historical flood forecasting information and historical river basin environment data; determining the case similarity between the historical river basin environment data of each river basin flood case and the river basin environment data, and determining the river basin flood case corresponding to the case similarity higher than a preset threshold as a target case; determining the similarity weight and the confidence weight of each target case; for each target case, correcting the historical flood forecasting information thereof based on the similarity weight and the confidence weight, to obtain the corrected forecasting information of each target case; and determining the flood forecasting information of the to-be-predicted river basin based on the plurality of corrected forecasting information. The above scheme can improve the accuracy of flood forecasting.

Description

Technical Field

[0001] This invention relates to the field of flood forecasting technology, and in particular to a flood forecasting method, apparatus, storage medium, and computer based on historical cases. Background Technology

[0002] Flood forecasting is a core component of flood control and disaster reduction efforts, and its accuracy directly impacts the scientific rigor and timeliness of flood control decision-making. Currently, case-based reasoning (CBR) flood forecasting technology has become an important research direction in this field due to its ability to fully utilize historical flood experience data.

[0003] Currently, the core logic of flood forecasting based on case-based reasoning is to construct a historical case database. Based on the basin environmental data such as rainfall, total rainfall, rainfall pattern, and rising water level of the current basin to be predicted, the database is used to select the basin flood case that is closest to the basin environmental data of the current basin to be predicted. The historical flood forecast information corresponding to this basin flood case is then used as the flood forecast information for the basin to be predicted, thus achieving flood forecasting for the basin to be predicted.

[0004] However, existing CBR-type flood forecasting schemes still have significant shortcomings in practical applications. They rely on historical flood forecast information corresponding to a single basin flood case as the flood forecast information for the basin to be predicted. Due to the occasional errors in the forecast information corresponding to the basin flood case, the current flood forecast may be biased, resulting in low accuracy of existing flood forecasting schemes. Summary of the Invention

[0005] In view of this, this application provides a flood forecasting method, apparatus, storage medium and computer based on historical cases, with the main purpose of solving the technical problem of low accuracy of existing flood forecasting schemes.

[0006] According to a first aspect of the present invention, a flood forecasting method based on historical cases is provided, the method comprising: Obtain flood prediction data for the watershed to be predicted, wherein the flood prediction data includes watershed type data, flood type data, and watershed environmental data; Based on the watershed type data and the flood type data, multiple watershed flood cases are obtained from a preset case library, wherein the watershed flood cases include historical flood forecast information and historical watershed environmental data; For each watershed flood case, the case similarity between the historical watershed environmental data of the watershed flood case and the watershed environmental data is determined to obtain the case similarity corresponding to each watershed flood case, and the watershed flood cases corresponding to the case similarity higher than the preset similarity threshold are determined as target cases; Based on the case similarity of each target case, a similarity weight is determined for each target case, and based on the confidence level of the historical flood forecast information of each target case, a confidence weight is determined for each target case. For each target case, the historical flood forecast information of the target case is corrected based on the similarity weight and confidence weight of the target case to obtain the corrected forecast information of each target case; Based on the revised forecast information for each target case, the flood forecast information for the watershed to be predicted is determined.

[0007] In an optional embodiment, the watershed environmental data includes real-time meteorological data, basic geographic data, vegetation distribution data, seasonal data, water conservancy project data, human activity data, and flood change data; the historical watershed environmental data includes historical meteorological data, comparative geographic data, comparative vegetation data, comparative seasonal data, comparative water conservancy data, comparative human activity data, and comparative flood change data; determining the case similarity between the historical watershed environmental data and the watershed environmental data for the watershed flood case includes: calculating a first parameter similarity between the real-time meteorological data and the historical meteorological data, and a second parameter similarity between the basic geographic data and the comparative geographic data. The similarity scores are calculated as follows: numerical similarity, a third parameter similarity between the vegetation distribution data and the control vegetation data, a fourth parameter similarity between the seasonal data and the control seasonal data, a fifth parameter similarity between the water conservancy project data and the control water conservancy data, a sixth parameter similarity between the human activity data and the control human activity data, and a seventh parameter similarity between the flood change data and the control flood change data. Based on these first, second, third, fourth, fifth, sixth, and seventh parameter similarities, the case similarity score is generated.

[0008] In an optional embodiment, the basic geographic data includes soil moisture content; the method for determining the similarity threshold includes: comparing the soil moisture content with multiple preset moisture content intervals, wherein each moisture content interval corresponds to a preset similarity threshold; determining the moisture content interval in which the soil moisture content is located, and determining the preset similarity threshold corresponding to the moisture content interval as the similarity threshold.

[0009] In an optional embodiment, determining the similarity weight of each target case based on the case similarity of each target case includes: calculating the sum of the case similarities of all target cases to obtain a total similarity; and calculating the quotient of the case similarity of each target case and the total similarity to obtain the similarity weight of the target case.

[0010] In an optional embodiment, determining the confidence weight of each target case based on the confidence level of historical flood forecast information for each target case includes: obtaining historical actual flood information corresponding to each target case, and calculating the forecast confidence level of the target case based on the following formula: Y = (1-| - | / ) × 100% Where Y is the forecast confidence level. Historical flood forecast information, The data is based on historical flood information. The sum of the forecast confidence scores of all target cases is calculated to obtain the total confidence score. The quotient of the forecast confidence score of each target case and the total confidence score is calculated to obtain the confidence weight of each target case.

[0011] In an optional embodiment, the step of correcting the historical flood forecast information of the target case based on the similarity weight and confidence weight to obtain the corrected forecast information of each target case includes: calculating the product of the historical flood forecast information, similarity weight and confidence weight of the target case to obtain the corrected forecast information of the target case.

[0012] In an optional embodiment, after determining the flood forecast information of the watershed to be predicted based on the corrected forecast information of each target case, the method further includes: obtaining a forecast confidence score of the flood forecast information based on the forecast confidence of each target case and the difference between the historical actual flood information and the historical flood forecast information of each target case.

[0013] According to a second aspect of the present invention, a flood forecasting device based on historical cases is provided, the device comprising: The parameter acquisition module is used to acquire flood prediction data for the watershed to be predicted, wherein the flood prediction data includes watershed type data, flood type data, and watershed environmental data. The case filtering module is used to obtain multiple watershed flood cases from a preset case library based on the watershed type data and the flood type data. The watershed flood cases include historical flood forecast information and historical watershed environmental data. The similarity calculation module is used to determine the case similarity between the historical watershed environmental data and the watershed environmental data of each watershed flood case, obtain the case similarity corresponding to each watershed flood case, and determine the watershed flood cases corresponding to the case similarity higher than the preset similarity threshold as target cases; The weight calculation module is used to determine the similarity weight of each target case based on the case similarity of each target case, and to determine the confidence weight of each target case based on the confidence of the historical flood forecast information of each target case; The forecast correction module is used to correct the historical flood forecast information of each target case based on the similarity weight and confidence weight of the target case, so as to obtain the corrected forecast information of each target case. The result output module is used to determine the flood forecast information of the watershed to be predicted based on the corrected forecast information of each target case.

[0014] According to a third aspect of the invention, a storage medium is provided having a computer program stored thereon, which, when executed by a processor, implements the above-described flood forecasting method based on historical cases.

[0015] According to a fourth aspect of the present invention, a computer is provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor, when executing the program, implements the above-described flood forecasting method based on historical cases.

[0016] This invention provides a flood forecasting method, apparatus, storage medium, and computer based on historical cases. By selecting multiple flood cases from a case library based on the basin type and flood type of the watershed to be predicted, it avoids the shortcomings of existing technologies that rely on single cases and are susceptible to sporadic errors. Furthermore, by calculating the similarity of cases to select target cases, and combining similarity weights and confidence weights to correct the historical flood forecasting information of the target cases, it fully integrates the effective information of multiple highly similar and reliable cases, reduces the interference of single case errors on the forecast results, and thus significantly improves the accuracy of flood forecasting.

[0017] The above description is only an overview of the technical solution of this application. In order to better understand the technical means of this application and to implement it in accordance with the contents of the specification, and to make the above and other objects, features and advantages of this application more obvious and understandable, the following are specific embodiments of this application. Attached Figure Description

[0018] The accompanying drawings, which are included to provide a further understanding of the invention and form part of this application, illustrate exemplary embodiments of the invention and, together with their description, serve to explain the invention and do not constitute an undue limitation thereof. In the drawings: Figure 1 A flowchart illustrating a flood forecasting method based on historical cases provided by an embodiment of the present invention is shown. Figure 2 A schematic diagram of the structure of a flood forecasting device based on historical cases provided in an embodiment of the present invention is shown; Figure 3 The diagram shows a schematic of the structure of a computer that performs a flood forecasting method based on historical cases, according to an embodiment of the present invention. Detailed Implementation

[0019] The present invention will be described in detail below with reference to the accompanying drawings and embodiments. It should be noted that, unless otherwise specified, the embodiments and features described in the present application can be combined with each other.

[0020] Currently, the core logic of case-based reasoning (CBR) flood forecasting is to construct a historical case database. Based on watershed environmental data such as rainfall, total precipitation, rainfall pattern, and initial water level of the current watershed to be predicted, the database is used to select the watershed flood case that most closely matches the current watershed's environmental data. The historical flood forecast information corresponding to this case is then used as the flood forecast information for the watershed to be predicted. However, existing CBR-based flood forecasting schemes still have significant shortcomings in practical applications. Relying on historical flood forecast information corresponding to a single watershed flood case as the flood forecast information for the watershed to be predicted is susceptible to occasional errors in the forecast information corresponding to the watershed flood case, leading to deviations in the current flood forecast and consequently lower accuracy.

[0021] To address the above problems, in one embodiment, such as Figure 1 As shown, a flood forecasting method based on historical cases is provided. Taking the application of this method to a computer as an example, the method includes the following steps: 101. Obtain flood prediction data for the watershed to be predicted, wherein the flood prediction data includes watershed type data, flood type data, and watershed environmental data.

[0022] Here, the watershed to be predicted refers to a specific river basin for which flood forecasting needs to be conducted. Watershed type data includes the watershed type itself, such as mountainous basins, plains basins, and coastal basins. Furthermore, watershed environmental data can include real-time meteorological data, basic geographic data, vegetation distribution data, seasonal data, hydraulic engineering data, human activity data, and flood change data. Additionally, flood type data refers to the causal type of floods occurring in the watershed to be predicted, specifically including rainstorm floods, snowmelt floods, and typhoon floods, to clearly distinguish the triggering sources of different floods. Here, by combining the geographical characteristics of the watershed to be predicted (such as whether it is near the sea or whether there is snow accumulation on high mountains), real-time meteorological data (intensity and duration of rainstorms, typhoon paths and impact range), hydrological monitoring data (snowmelt rate, source of river water), and historical flood cause records, relevant personnel can determine the flood type data of possible floods in the watershed to be predicted by analyzing the dominant inducing factors. Among them, floods mainly induced by rainstorms are rainstorm floods, floods mainly induced by snowmelt are snowmelt floods, and floods mainly induced by typhoon weather systems are typhoon floods.

[0023] Specifically, real-time meteorological data includes rainfall (in millimeters), rainfall intensity (in millimeters per hour), rainfall duration (in hours), temperature (in degrees Celsius), and relative humidity.

[0024] Further, the basic geographic data include: average slope of the watershed, average elevation of the watershed, soil moisture content, watershed morphology coefficient (ratio of watershed length to watershed width), soil type, and groundwater depth (unit: meters, used to characterize the average depth of the watershed). Furthermore, vegetation distribution data includes: vegetation coverage, vegetation type (including: broad-leaved forest and coniferous forest, grassland, farmland, desert), growth stage (including: seedling stage, growing season, maturity stage, and withering stage), and vegetation transpiration coefficient (used to characterize the vegetation's water consumption capacity). Furthermore, seasonal data includes: flood season / non-flood season parameters, and quarter divisions (including: spring, summer, autumn, and winter). Furthermore, water conservancy project data includes: the number of reservoirs in the basin, the reservoir capacity, the flood control standard of dikes, the distribution density of water conservancy facilities (unit: facilities / 100km²), and the degree of river dredging (including no dredging, light dredging, moderate dredging, and deep dredging). Further, human activity data include: urbanization rate (used to characterize the proportion of urban built-up area in the basin), cultivated land ratio (used to characterize the proportion of cultivated land area in the basin), population density, distribution of major transportation routes (unit: km / 100km), and irrigated area ratio (used to characterize the proportion of irrigated farmland area in the basin).

[0025] Further flood change data include: peak flood value, flood duration, rate of rise, rate of receding water, and total flood volume.

[0026] Here, the aforementioned information can be obtained continuously and in real time from the host computer. Specifically, a multi-source data acquisition module can be built on the server or other host computer to synchronously acquire the aforementioned watershed environmental data of the target watershed: First, in terms of meteorological data, real-time rainfall monitoring data from meteorological radar, temperature / humidity data obtained from satellite remote sensing, and observation data from ground meteorological stations can be integrated; second, in terms of hydrological data, data such as soil moisture content, initial river flow, and groundwater depth can be collected through IoT sensors, combined with real-time monitoring of rising water levels from hydrological stations; furthermore, in terms of basic data, real-time vegetation coverage and topographic data can be obtained by calling the Geographic Information System (GIS) interface, and urbanization rate and the latest operational data of water conservancy projects can be obtained through the government data sharing platform.

[0027] Furthermore, the collected watershed environmental data can be fused: First, the Kalman filter algorithm is used to eliminate data noise, where the state equation of the Kalman filter algorithm is shown in Equation 1: (1) in, N Representing a moment, This refers to the current state of a watershed environmental data point, i.e., the watershed environmental data at the current moment. A Here is the state transition matrix. B To control the input matrix, This refers to the state at the previous moment, i.e., the watershed environmental data at the previous moment. The state input vector, i.e., the first... N 1 External control factors that constantly exert an active influence on the hydrological state of a watershed, namely, human or natural intervention / driving variables. This is the process noise vector.

[0028] Furthermore, a correlation between the actual state and the observed data can be established through observation equations. This allows for the comparison of the estimated actual state of the watershed environment data with the actual multi-source data collected, thus correcting errors. Here, Equation 2 can be used to correct errors in the watershed environment data: =H + (2) in, N Representing a moment, This represents the current state of a watershed environmental data point. The revised watershed environmental data, H For the observation matrix, V The noise is Gaussian white noise. Here, for each watershed environmental data, contradictory data is eliminated through data consistency checks (error ≤ 5% is acceptable). Then, missing data is filled with complementary data from multiple sources. For example, when hydrological station data is missing, IoT sensor data and satellite remote sensing data are fused and estimated to ensure data reliability.

[0029] Furthermore, each piece of watershed environmental data is categorized and processed, converting non-digital parameters such as vegetation type, soil type, and soil type into digital codes. Specifically, for soil type, "sandy soil" is coded as 1, "sandy soil" as 2, "clay soil" as 3, and "mixed soil" as 4.

[0030] Similarly, for vegetation type, "broadleaf forest" is coded as 1, "coniferous forest" as 2, "grassland" as 3, "farmland" as 4, and "desert" as 5; further, for growth stage, "seedling stage" is coded as 1, "growing season" as 2, "maturity stage" as 3, and "wilt stage" as 4; further, for flood season / non-flood season parameters, May-October in the Northern Hemisphere is coded as 1, and so on. The remaining months in the Southern Hemisphere are coded as 0, November to April in the Southern Hemisphere are coded as 1, and the remaining months in the Southern Hemisphere are coded as 0. Furthermore, in terms of seasons, "Spring" is coded as 1, "Summer" as 2, "Autumn" as 3, and "Winter" as 4. Furthermore, in terms of the degree of river dredging, "No dredging" is coded as 1, "Light dredging" as 2, "Moderate dredging" as 3, and "Deep dredging" as 4.

[0031] Furthermore, all watershed environmental data can be uniformly mapped to the [0, 1] interval to form a standardized case parameter vector.

[0032] 102. Based on the watershed type data and the flood type data, obtain multiple watershed flood cases from a preset case library, wherein the watershed flood cases include historical flood forecast information and historical watershed environmental data.

[0033] Among them, historical flood forecast information refers to the flood forecast records made based on the historical watershed environmental data at the time when the flood actually occurred in the target case. The core includes the historical forecast flood peak (unit: m³ / s), the historical forecast flood occurrence time and the historical forecast receding time. It is the original forecast data corresponding to the historical actual flood information of the case.

[0034] Furthermore, historical watershed environmental data includes historical meteorological data, comparative geographical data, comparative vegetation data, comparative seasonal data, comparative water conservancy data, comparative human activity data, and comparative flood change data when the flood actually occurred in the watershed in history.

[0035] Here, in the pre-set case library, all basin flood cases have been structurally archived according to the two-dimensional classification rule of "basin type - flood type". That is, each flood case has corresponding basin type data and flood type data. At the same time, each case is associated with complete historical basin environmental data and corresponding historical flood forecast information. The historical flood forecast information includes core forecast results such as flood peak value, flood peak time, and receding time.

[0036] Specifically, based on the basin type data (such as mountain basins, plain basins, coastal basins, etc.) and flood type data (such as rainstorm floods, snowmelt floods, typhoon floods, etc.) of the basin to be predicted, the case library's classification indexing mechanism is used for precise screening. Multiple basin flood cases that are consistent with the basin and flood type of the basin to be predicted are matched and obtained from the case library, providing basic data support for subsequent case similarity calculation and forecast information correction.

[0037] Furthermore, when generating watershed flood cases for the case library, the first step is to identify floods that have occurred in different watersheds globally over the past 50 years and obtain historical flood process data for each flood. Each flood is then designated as a watershed flood case, and its historical flood process data is collected and linked to the case. This historical flood process data covers 32 core parameters across 7 categories related to the occurrence of floods in the watershed, specifically including: Five historical meteorological data items were collected, including: rainfall, rainfall intensity, rainfall duration, temperature, and relative humidity. Compare the five flood change data points, including: peak flood value, flood duration, flood rise rate, flood receding rate, and total flood volume; The data were compared against six geographical data points, including: average slope of the watershed, average elevation of the watershed, soil moisture content, watershed morphology coefficient, soil type, and groundwater depth. Four vegetation data points were compared: vegetation coverage, vegetation type, growth stage, and vegetation transpiration coefficient. Compare two seasonal data items, including: flood season / non-flood season and quarterly division; Five water conservancy data points were compared, including: the number of reservoirs, reservoir capacity, flood control standards of dikes and dams, distribution density of water conservancy facilities, and the degree of river dredging. Five data points related to human activities were compared: urbanization rate, proportion of arable land, population density (unit: people / km²), distribution of major transportation routes, and proportion of irrigated area.

[0038] Furthermore, for each river basin flood case, the aforementioned 32 core parameters across 7 categories are linked to it.

[0039] Furthermore, the historical watershed environmental data collected for the aforementioned watershed flood cases can be preprocessed. Here, the 3σ criterion can be used to remove outliers. Specifically, for each category of historical watershed environmental data, if the parameter value of a certain historical watershed environmental data point deviates from ±3 times the standard deviation of the mean of that parameter in that category of data, then the historical watershed environmental data with significantly deviated parameters is removed. Further, the K-nearest neighbor interpolation method (K=5) can be used to impute missing values. If a certain historical watershed environmental data point for a watershed flood case is missing, then the five watershed flood cases of the same watershed type and flood type as the case, and which are most similar in other parameters, are selected from the case database. The mean of the corresponding parameter in these five cases is used as the imputation value for the missing data, thereby ensuring the integrity and consistency of the data.

[0040] Furthermore, each piece of historical watershed environmental data is categorized and processed: non-digital parameters such as vegetation type, soil type, and soil type are converted into digital codes. Specifically, for soil type, "sandy soil" is coded as 1, "sandy soil" as 2, "clay soil" as 3, and "mixed soil" as 4. Similarly, for vegetation type, "broadleaf forest" is coded as 1, "coniferous forest" as 2, "grassland" as 3, "farmland" as 4, and "desert" as 5; further, for growth stage, "seedling stage" is coded as 1, "growing season" as 2, "maturity stage" as 3, and "withering stage" as 4; further, for flood season / non-flood season, May to October in the Northern Hemisphere is coded as 1, and the rest of the months in the Northern Hemisphere is coded as 0, while November to April in the Southern Hemisphere is coded as 1, and the rest of the months in the Southern Hemisphere is coded as 0; further, for season division, "spring" is coded as 1, "summer" as 2, "autumn" as 3, and "winter" as 4; further, for river dredging degree, "undredged" is coded as 1, "lightly dredged" as 2, "moderately dredged" as 3, and "deeply dredged" as 4.

[0041] Furthermore, all historical watershed environmental data can be uniformly mapped to the [0, 1] interval to form a standardized case parameter vector. Specifically, for meteorological parameters (such as rainfall, rainfall intensity, etc.), the Z-score algorithm can be used for normalization, and the normalization formula is shown in Formula 3: Y=(Q-μ) / σ (3) Where Y is the normalized historical watershed environmental data in the watershed flood case, Q is the historical watershed environmental data before normalization in the watershed flood case, μ is the mean of this type of historical watershed environmental data in all watershed flood cases, and σ is the standard deviation of this type of historical watershed environmental data in all watershed flood cases. Furthermore, for geographic parameters (such as slope, altitude, etc.), the min-max algorithm can be used for normalization, and the normalization formula is shown in Formula 4: Y = (Q - Qmin) / (Qmax - Qmin) (4) Where Y represents the normalized historical watershed environmental data in the watershed flood case, Q represents the historical watershed environmental data before normalization in the watershed flood case, Qmin represents the minimum value of this type of historical watershed environmental data in all watershed flood cases, and Qmax represents the maximum value of this type of historical watershed environmental data in all watershed flood cases.

[0042] Furthermore, parameters related to human activities (such as urbanization rate and arable land ratio) are normalized using the entropy method, based on formulas 5 to 8: (5) (6) (7) (8) in, Let be the proportion of the i-th sample in the j-th human activity category parameter. Let be the original data value of the i-th sample in the j-th parameter. Let be the entropy value of the j-th parameter, reflecting the degree of information dispersion of that parameter, and k be the entropy calculation coefficient. Let be the difference coefficient of the j-th parameter, reflecting the degree of influence of this parameter on the decision. This represents the normalized weight of the j-th parameter. Here, the proportion, entropy value, and difference coefficient of the j-th historical watershed environmental data in the case of watershed flood are calculated sequentially to obtain its normalized weight. This method can effectively preserve the original distribution characteristics of the data while eliminating the dimensional differences between parameters.

[0043] Finally, the archiving is completed according to the two-dimensional classification rule of "basin type - flood cause" (basin type includes mountain basin, plain basin, and coastal basin; flood cause includes rainstorm flood, snowmelt flood, and typhoon flood), forming a structured and scalable case library. Each basin flood case is associated with a complete correspondence between "32 parameter sets and flood forecast information (peak flood value, occurrence time, and receding time)", so that multiple suitable basin flood cases can be quickly matched and obtained based on the basin type data and flood type data of the basin to be predicted.

[0044] Furthermore, for each type of historical watershed environmental data, the contribution of each type of historical watershed environmental data to historical flood forecasting information can be determined based on a random forest regression model, thus determining the contribution weight of each type of historical watershed environmental data. Specifically, typical watershed flood cases corresponding to each watershed type and flood cause can be extracted from the case library (no less than 1000 samples per type). The flood forecast results of each case (including peak flood value, flood occurrence time, and receding time) are used as the dependent variable, and the aforementioned 32 core parameters across 7 categories are used as independent variables to construct a random forest regression model. The specific model parameters are set as follows: 500 decision trees, maximum depth 15, minimum number of sample splits 10, and feature sampling ratio 0.8. Through the model training process, the feature importance score of each type of historical watershed environmental data to historical flood forecasting information is calculated. A higher score indicates a greater influence of that type of parameter on historical flood forecasting information, providing a basis for parameter weight allocation in subsequent similarity calculations.

[0045] 103. For each of the aforementioned watershed flood cases, determine the case similarity between the historical watershed environmental data of the watershed flood case and the watershed environmental data, obtain the case similarity corresponding to each of the aforementioned watershed flood cases, and determine the watershed flood cases corresponding to the case similarity higher than the preset similarity threshold as target cases.

[0046] Here, for each of the aforementioned watershed flood cases, the following parameters can be calculated: a first parameter similarity between real-time meteorological data and historical meteorological data corresponding to the watershed flood case; a second parameter similarity between the basic geographic data and the control geographic data; a third parameter similarity between the vegetation distribution data and the control vegetation data; a fourth parameter similarity between the seasonal data and the control seasonal data; a fifth parameter similarity between the water conservancy project data and the control water conservancy data; a sixth parameter similarity between the human activity data and the control human activity data; and a seventh parameter similarity between the flood change data and the control flood change data. Specifically, Python data analysis tools (such as NumPy, Pandas, etc.) can be used to vectorize the environmental data of each watershed to be predicted, obtaining the real-time parameter vector corresponding to each watershed environmental data. All historical watershed environmental data of each watershed flood case can be vectorized to obtain the case parameter vector corresponding to each historical watershed environmental data, so as to perform subsequent similarity calculations.

[0047] Furthermore, regarding the first parameter similarity between real-time meteorological data and historical meteorological data, the cosine similarity between the rainfall, rainfall intensity, rainfall duration, temperature, and relative humidity in the real-time meteorological data and the corresponding parameters in the historical meteorological data can be calculated. For example, the cosine similarity between the real-time parameter vector corresponding to the rainfall in the real-time meteorological data and the case parameter vector corresponding to the rainfall in the historical meteorological data can be calculated, and the cosine similarity between the real-time parameter vector corresponding to the rainfall intensity in the real-time meteorological data and the case parameter vector corresponding to the rainfall intensity in the historical meteorological data can be calculated, until the cosine similarity calculation for each parameter in the real-time meteorological data and the historical meteorological data is completed. Furthermore, the obtained cosine similarities can be accumulated and averaged or otherwise normalized to obtain the first parameter similarity.

[0048] Furthermore, the calculation process for the second, third, fourth, fifth, sixth, and seventh parameter similarities can be referenced from the calculation process for the first parameter similarity, and will not be elaborated here.

[0049] Furthermore, the case similarity is generated based on the first parameter similarity, the second parameter similarity, the third parameter similarity, the fourth parameter similarity, the fifth parameter similarity, the sixth parameter similarity, and the seventh parameter similarity. Specifically, the first parameter similarity, the second parameter similarity, the third parameter similarity, the fourth parameter similarity, the fifth parameter similarity, the sixth parameter similarity, and the seventh parameter similarity can be normalized by averaging to obtain the case similarity of the flood cases in the basin. Further, the case similarity corresponding to each flood case in the basin can be calculated using the above method, and the case similarity corresponding to each flood case in the basin can be compared with a preset similarity threshold. The flood cases in the basin corresponding to case similarities higher than the similarity threshold are identified as target cases. Here, the value of the similarity threshold can be determined based on the actual situation.

[0050] In the actual calculation process, the contribution weight of each type of historical watershed environmental data can be determined. When calculating the cosine similarity between historical watershed environmental data and a watershed environmental data, the contribution weight of the historical watershed environmental data is multiplied by the calculated cosine similarity to obtain the corrected cosine similarity, and then subsequent calculations are performed.

[0051] Furthermore, the growth stage, flood season / non-flood season parameters, and soil moisture content in the watershed environmental data can be obtained to determine the watershed scenario. When a watershed scenario is identified, the contribution weights of rainfall, groundwater depth, and soil type in the historical watershed environmental data are adjusted accordingly. The method for determining the watershed scenario is shown in Table 1.

[0052] Table 1 Here, if the growth stage in the watershed environmental data is the growing season, the flood season / non-flood season parameter is the flood season, and the soil moisture content reaches the preset saturation threshold (60%), then the watershed scenario is determined as Scenario 1. In this case, the contribution weight of rainfall in the historical watershed environmental data is increased by 30%, and the contribution weight of rainfall intensity is increased by 30%. Similarly, if the growth stage in the watershed environmental data is the mature stage, the flood season / non-flood season parameter is the non-flood season, and the soil moisture content has not reached the preset saturation threshold (60%), then the watershed scenario is determined as Scenario 8. In this case, the contribution weight of soil type is increased by 25%, and the contribution weight of groundwater depth is increased by 25%. Here, each watershed scenario corresponds to one or more types of historical watershed environmental data, and the adjustment method for the contribution weight of each type of historical watershed environmental data is as follows. The specific adjustment algorithm and adjustment value can be determined according to the actual situation and are not limited here.

[0053] Furthermore, methods for determining the similarity threshold include: First, the soil moisture content in the watershed environmental data is compared with multiple preset moisture content intervals. Each moisture content interval corresponds to a different numerical range and has a different preset similarity threshold. The preset similarity thresholds for each moisture content interval are different. Then, the moisture content interval in which the soil moisture content value is located is determined, and the preset similarity threshold corresponding to the moisture content interval is determined as the similarity threshold.

[0054] As an example, based on the soil saturation quantification standard, soil moisture content ≥80% is considered saturated, corresponding to the saturated moisture content range. Soil moisture content between 60% and 80% is considered semi-saturated, corresponding to the semi-saturated moisture content range. Soil moisture content <60% is considered arid, corresponding to the arid moisture content range. Furthermore, the preset similarity threshold for saturated state is 0.75, for semi-saturated state it is 0.70, and for arid state it is 0.65; these thresholds are dynamically updated based on real-time soil moisture content, with an update frequency of once per hour. Furthermore, if the soil moisture content in the watershed environmental data is 50%, corresponding to the arid moisture content range, 0.65 is determined as the similarity threshold.

[0055] 104. Based on the case similarity of each target case, determine the similarity weight of each target case, and based on the confidence of the historical flood forecast information of each target case, determine the confidence weight of each target case.

[0056] Specifically, the sum of the case similarities of all the target cases can be calculated to obtain the total similarity; further, the case similarity of the target cases can be divided by the total similarity to obtain the similarity weight of the target cases.

[0057] Furthermore, historical flood information corresponding to each target case is obtained. Here, historical flood information refers to the actual flood monitoring data of the river basin corresponding to the target case during the historical basin environmental data collection period when floods actually occurred. The core data includes the actual flood peak value, the actual flood occurrence time, and the actual receding time, which are key references for evaluating the accuracy of the historical flood forecast information of the target case. Furthermore, the forecast confidence level of the target case is calculated based on the following formula 9: Y = (1-| - | / )×100%(9) Where Y is the forecast confidence level. Historical flood forecast information, The system uses historical flood information; further, it calculates the sum of the forecast confidence scores for each target case to obtain the total confidence score; then, it divides the forecast confidence score of each target case by the total confidence score to obtain the confidence weight of that target case. Based on the above method, the confidence weight of each target case can be calculated.

[0058] 105. For each target case, based on the similarity weight and confidence weight of the target case, the historical flood forecast information of the target case is corrected to obtain the corrected forecast information of each target case.

[0059] Specifically, the revised forecast information for the target case can be obtained by multiplying its historical flood forecast information, similarity weight, and confidence weight. Here, the revised forecast information may include revised values ​​for the predicted flood peak, predicted flood occurrence time, and predicted receding time; wherein, the revised value for the predicted flood peak of the target case is the product of the historical predicted flood peak value, similarity weight, and confidence weight; the revised value for the predicted flood occurrence time of the target case is the product of the historical predicted flood occurrence time, similarity weight, and confidence weight; and the revised value for the predicted receding time of the target case is the product of the historical predicted receding time, similarity weight, and confidence weight.

[0060] 106. Based on the revised forecast information of each target case, determine the flood forecast information of the watershed to be predicted.

[0061] Here, the forecast peak correction values ​​for each target case can be summed and averaged to obtain the peak flood forecast information for the basin to be predicted; the forecast flood occurrence time correction values ​​for each target case can be summed and averaged to obtain the flood occurrence time forecast information for the basin to be predicted; the forecast receding time correction values ​​for each target case can be summed and averaged to obtain the receding time forecast information for the basin to be predicted; furthermore, the peak flood forecast information, the flood occurrence time forecast information, and the receding time forecast information are used as the flood forecast information for the basin to be predicted.

[0062] The flood forecasting method based on historical cases provided in this embodiment selects multiple flood cases from a case library based on the basin type and flood type of the basin to be predicted, avoiding the shortcomings of existing technologies that rely on a single case and are susceptible to accidental errors. Furthermore, it selects target cases by calculating case similarity and corrects the historical flood forecasting information of the target cases by combining similarity weight and confidence weight. This fully integrates the effective information of multiple highly similar and reliable cases, reduces the interference of single case errors on the forecast results, and thus significantly improves the accuracy of flood forecasting.

[0063] In an optional embodiment, after obtaining the flood forecast information of the watershed to be predicted by weighted summation of the historical flood forecast information of each target case based on the forecast weight of each target case, the method further includes: Based on the forecast confidence level of each target case and the difference between the historical actual flood information and the historical flood forecast information of each target case, the forecast confidence score of the flood forecast information is obtained.

[0064] Specifically, for each target case, the difference between its historical flood forecast information and historical actual flood information is calculated. That is, the difference between the historical forecast flood peak and the actual flood peak of the target case is calculated, the first time difference between the historical forecast flood occurrence time and the actual flood occurrence time of the target case is calculated, and the second time difference between the historical forecast receding time and the actual receding time of the target case is calculated.

[0065] Furthermore, the average confidence level of the forecasts for all target cases is calculated, which is the arithmetic mean of the confidence weights of all target cases. Further, the flood forecast information scores 30 points when the average peak difference of all target cases is less than 5%, 20 points when the average peak difference is between 5% and 8%, and 10 points when the average peak difference is between 8% and 10%. Furthermore, the flood forecast information scores 30 points when the average first time difference of all target cases is less than 1 hour, and 30 points when the average first time difference is between 1 and 2 hours. The flood forecast information scores 20 points; furthermore, if the mean of the second time difference for all target cases is less than 1 hour, the flood forecast information scores 30 points; if the mean of the second time difference for all target cases is between 1 and 2 hours, the flood forecast information scores 20 points; furthermore, if the mean confidence level is greater than 90%, the flood forecast information scores 40 points; if the mean confidence level is between 80% and 90%, the flood forecast information scores 30 points; and if the mean confidence level is between 70% and 80%, the flood forecast information scores 20 points.

[0066] Furthermore, the flood forecast information is weighted and summed based on the scores corresponding to the peak difference, the first time difference, the second time difference, and the mean confidence level to obtain a flood forecast reliability score. The flood forecast information and its reliability score are then sent to the host computer used by relevant personnel to obtain the final forecast result.

[0067] Furthermore, the system can monitor the actual flood development dynamics in the target basin in real time, comparing the actual flood data (including the actual peak flood value, the actual flood occurrence time, and the actual receding time) with the previous forecast results every 6 hours to accurately calculate various error values. If the comparison finds that the error between the forecasted peak flood value and the actual peak flood value is >10%, or the maximum of the deviations between the flood occurrence time and the receding time is >3 hours, the model update mechanism will be automatically triggered. Specifically, the current complete flood process data is first added to the case library and archived according to the two-dimensional classification rule of "basin type-flood cause". Then, the random forest model is retrained to calibrate the parameter weight matrix. At the same time, the relevant parameters are optimized based on the error analysis results: if the normalized error of a certain type of parameter accounts for more than 30% of the total error, the normalization coefficient of that type of parameter is adjusted; if the proportion of low confidence cases after threshold screening exceeds 20%, the screening threshold (±0.02) under the corresponding hydrological conditions is fine-tuned to form a closed-loop mechanism of "forecast-monitoring-feedback-optimization" to continuously improve the accuracy and reliability of flood forecasts.

[0068] The flood forecasting method based on historical cases provided in this embodiment performs case matching through multi-dimensional watershed environmental data and dynamically determines the similarity threshold by combining soil moisture content, thereby effectively improving the accuracy of case selection. Furthermore, the forecast information is corrected by using a dual weighting of similarity weight and confidence weight, which greatly reduces the error interference of a single case. At the same time, a forecast credibility score is generated to intuitively present the reliability of the forecast, significantly improving the accuracy and scientific nature of flood forecasts and providing solid and reliable data support for flood control scheduling decisions.

[0069] Furthermore, as Figure 1 The specific implementation of the method shown in this embodiment provides a flood forecasting device based on historical cases, such as... Figure 2 As shown, the device includes: a parameter acquisition module 21, a case screening module 22, a similarity calculation module 23, a weight calculation module 24, a prediction correction module 25, and a result output module 26.

[0070] The parameter acquisition module 21 can be used to acquire flood prediction data of the watershed to be predicted, wherein the flood prediction data includes watershed type data, flood type data and watershed environmental data; The case filtering module 22 can be used to obtain multiple watershed flood cases from a preset case library based on the watershed type data and the flood type data, wherein the watershed flood cases include historical flood forecast information and historical watershed environmental data; The similarity calculation module 23 can be used to determine the case similarity between the historical watershed environmental data and the watershed environmental data of each watershed flood case, obtain the case similarity corresponding to each watershed flood case, and determine the watershed flood cases corresponding to the case similarity higher than the preset similarity threshold as target cases; The weight calculation module 24 can be used to determine the similarity weight of each target case based on the case similarity of each target case, and to determine the confidence weight of each target case based on the confidence of the historical flood forecast information of each target case; The forecast correction module 25 can be used to correct the historical flood forecast information of each target case based on the similarity weight and confidence weight of the target case, so as to obtain the corrected forecast information of each target case. The result output module 26 can be used to determine the flood forecast information of the watershed to be predicted based on the corrected forecast information of each target case.

[0071] In specific application scenarios, the watershed environmental data includes real-time meteorological data, basic geographic data, vegetation distribution data, seasonal data, water conservancy project data, human activity data, and flood change data; the historical watershed environmental data includes historical meteorological data, comparative geographic data, comparative vegetation data, comparative seasonal data, comparative water conservancy data, comparative human activity data, and comparative flood change data; the similarity calculation module 23 can be specifically used to calculate the first parameter similarity between the real-time meteorological data and the historical meteorological data, the second parameter similarity between the basic geographic data and the comparative geographic data, and the vegetation distribution data. The case similarity is generated based on the third parameter similarity between the data and the control vegetation data, the fourth parameter similarity between the seasonal data and the control seasonal data, the fifth parameter similarity between the water conservancy project data and the control water conservancy data, the sixth parameter similarity between the human activity data and the control human activity data, and the seventh parameter similarity between the flood change data and the control flood change data; the case similarity is generated based on the first parameter similarity, the second parameter similarity, the third parameter similarity, the fourth parameter similarity, the fifth parameter similarity, the sixth parameter similarity, and the seventh parameter similarity.

[0072] In specific application scenarios, the weight calculation module 24 can be used to calculate the sum of case similarities of all the target cases to obtain the total similarity; and to calculate the quotient of the case similarity of the target case and the total similarity to obtain the similarity weight of the target case.

[0073] In specific application scenarios, the weight calculation module 24 can be used to obtain historical actual flood information corresponding to each target case, and calculate the forecast confidence of the target case based on the following formula: Y = (1-| - | / ) × 100% Where Y is the forecast confidence level. Historical flood forecast information, The data is based on historical flood information. The sum of the forecast confidence scores of all target cases is calculated to obtain the total confidence score. The quotient of the forecast confidence score of each target case and the total confidence score is calculated to obtain the confidence weight of each target case.

[0074] In specific application scenarios, the forecast correction module 25 can be used to calculate the product of the historical flood forecast information, similarity weight, and confidence weight of the target case to obtain the corrected forecast information of the target case.

[0075] In specific application scenarios, the forecast correction module 25 can also be used to obtain a forecast credibility score of the flood forecast information based on the forecast confidence of each target case and the difference between the historical actual flood information and the historical flood forecast information of each target case.

[0076] It should be noted that other corresponding descriptions of the functional units involved in the flood forecasting device based on historical cases provided in this embodiment can be found in [reference needed]. Figure 1 The corresponding description in [the document] will not be repeated here.

[0077] Based on the above, Figure 1 Accordingly, this embodiment also provides a storage medium storing a computer program that, when executed by a processor, implements the above-described method. Figure 1 The flood forecasting method shown is based on historical cases.

[0078] Based on this understanding, the technical solution of this application can be embodied in the form of a software product. The software product to be identified can be stored in a non-volatile storage medium (such as a CD-ROM, USB flash drive, mobile hard drive, etc.) and includes several instructions to cause a computer (such as a personal computer, server, or network device, etc.) to execute the methods described in the various implementation scenarios of this application.

[0079] Based on the above, Figure 1 The method shown, and Figure 2 The illustrated embodiment of a flood forecasting device based on historical cases is designed to achieve the aforementioned objectives, such as... Figure 3 As shown, this embodiment also provides a computer for executing a flood forecasting method based on historical cases. Specifically, it can be a personal computer, server, smartphone, tablet, smartwatch, or other network device. The computer includes a storage medium and a processor; the storage medium stores computer programs and an operating system; the processor executes the computer programs to implement the above-described... Figure 1 The method shown.

[0080] Optionally, the computer may also include internal memory, a communication interface, a network interface, a camera, radio frequency (RF) circuitry, sensors, audio circuitry, a Wi-Fi module, a display screen, and input devices such as a keyboard. The communication interface may also include a USB interface, a card reader interface, etc. The network interface may optionally include a standard wired interface or a wireless interface (such as a Wi-Fi interface).

[0081] Those skilled in the art will understand that the computer structure for recognizing operational actions provided in this embodiment does not constitute a limitation on the computer, and may include more or fewer components, or combine certain components, or have different component arrangements.

[0082] The storage medium may also include an operating system and a network communication module. The operating system is a program that manages the aforementioned computer hardware and software resources to be identified, supporting the operation of information processing programs and other software and / or programs to be identified. The network communication module is used to enable communication between the various components within the storage medium, as well as communication with other hardware and software in the information processing computer.

[0083] Through the above description of the embodiments, those skilled in the art can clearly understand that this application can be implemented by means of software plus necessary general-purpose hardware platforms, or it can be implemented by hardware. By applying the technical solution of this application, firstly, flood prediction data of the watershed to be predicted is obtained, wherein the flood prediction data includes watershed type data, flood type data, and watershed environmental data; then, based on the watershed type data and the flood type data, multiple watershed flood cases are obtained from a preset case library, wherein the watershed flood cases include historical flood forecast information and historical watershed environmental data; then, for each watershed flood case, the case similarity between the historical watershed environmental data of the watershed flood case and the case similarity of the watershed environmental data is determined, and the case similarity corresponding to each watershed flood case is obtained, and a similarity higher than the preset similarity is set. The watershed flood cases corresponding to the case similarity threshold are identified as target cases. Further, based on the case similarity of each target case, a similarity weight is determined for each target case, and based on the confidence level of the historical flood forecast information of each target case, a confidence weight is determined for each target case. Further, for each target case, based on the similarity weight and confidence weight, the historical flood forecast information of the target case is corrected to obtain corrected forecast information for each target case. Finally, based on the corrected forecast information of each target case, the flood forecast information for the watershed to be predicted is determined. Compared with existing technologies, this method can improve the accuracy of flood forecasting.

[0084] Those skilled in the art will understand that the accompanying drawings are merely schematic diagrams of a preferred embodiment, and the modules or processes shown in the drawings are not necessarily essential for implementing this application. Those skilled in the art will understand that the modules in the apparatus of the embodiment can be distributed within the apparatus of the embodiment as described, or can be modified to be located in one or more apparatuses different from this embodiment. The modules of the above-described embodiment can be combined into one module, or further divided into multiple sub-modules.

[0085] The serial numbers in this application are for descriptive purposes only and do not represent the superiority or inferiority of any particular implementation scenario. The above disclosures are merely a few specific implementation scenarios of this application; however, this application is not limited thereto, and any variations conceived by those skilled in the art should fall within the protection scope of this application.

Claims

1. A method of flood forecasting based on historical cases, characterized in that, The method includes: Obtain flood prediction data for the watershed to be predicted, wherein the flood prediction data includes watershed type data, flood type data, and watershed environmental data; Based on the watershed type data and the flood type data, multiple watershed flood cases are obtained from a preset case library, wherein the watershed flood cases include historical flood forecast information and historical watershed environmental data; For each watershed flood case, the case similarity between the historical watershed environmental data of the watershed flood case and the watershed environmental data is determined to obtain the case similarity corresponding to each watershed flood case, and the watershed flood cases corresponding to the case similarity higher than the preset similarity threshold are determined as target cases; Based on the case similarity of each target case, a similarity weight is determined for each target case, and based on the confidence level of the historical flood forecast information of each target case, a confidence weight is determined for each target case. For each target case, the historical flood forecast information of the target case is corrected based on the similarity weight and confidence weight of the target case to obtain the corrected forecast information of each target case; Based on the revised forecast information for each target case, the flood forecast information for the watershed to be predicted is determined.

2. The method of claim 1, wherein, The watershed environmental data includes real-time meteorological data, basic geographic data, vegetation distribution data, seasonal data, water conservancy project data, human activity data, and flood change data; The historical watershed environmental data includes historical meteorological data, comparative geographical data, comparative vegetation data, comparative seasonal data, comparative water conservancy data, comparative human activity data, and comparative flood change data; The determination of the similarity between the historical watershed environmental data and the watershed environmental data of the watershed flood case includes: The following parameters are calculated: the first parameter similarity between the real-time meteorological data and the historical meteorological data; the second parameter similarity between the basic geographic data and the control geographic data; the third parameter similarity between the vegetation distribution data and the control vegetation data; the fourth parameter similarity between the seasonal data and the control seasonal data; the fifth parameter similarity between the water conservancy project data and the control water conservancy data; the sixth parameter similarity between the human activity data and the control human activity data; and the seventh parameter similarity between the flood change data and the control flood change data. The case similarity is generated based on the first parameter similarity, the second parameter similarity, the third parameter similarity, the fourth parameter similarity, the fifth parameter similarity, the sixth parameter similarity, and the seventh parameter similarity.

3. The method of claim 2, wherein, The basic geographic data includes soil moisture content; the method for determining the similarity threshold includes: The soil moisture content is compared with multiple preset moisture content intervals, wherein each moisture content interval corresponds to a preset similarity threshold. The soil moisture content range is determined, and the preset similarity threshold corresponding to the moisture content range is determined as the similarity threshold.

4. The method of claim 1, wherein, The step of determining the similarity weight of each target case based on the case similarity of each target case includes: Calculate the sum of the case similarities of all the target cases to obtain the total similarity; The similarity weight of the target case is obtained by calculating the quotient of the case similarity of the target case and the sum of similarities.

5. The method of claim 4, wherein, The determination of the confidence weight for each target case based on the confidence level of historical flood forecast information for each target case includes: Obtain the historical actual flood information corresponding to each target case, and calculate the forecast confidence level of the target case based on the following formula: Y=（1-| - | / ）×100% wherein Y is a prediction confidence, is historical flood prediction information, is historical actual flood information; Calculate the sum of the prediction confidence scores for all the target cases to obtain the total confidence score; The confidence weight of the target case is obtained by calculating the quotient of the predicted confidence score of the target case and the total confidence score.

6. The method of claim 4, wherein, The historical flood forecast information of the target cases is corrected based on the similarity weight and confidence weight of the target cases to obtain the corrected forecast information for each target case, including: The corrected forecast information for the target case is obtained by multiplying the historical flood forecast information, similarity weight, and confidence weight of the target case.

7. The method of claim 5, wherein, After determining the flood forecast information for the watershed to be predicted based on the revised forecast information for each target case, the method further includes: Based on the forecast confidence level of each target case and the difference between the historical actual flood information and the historical flood forecast information of each target case, the forecast confidence score of the flood forecast information is obtained.

8. A flood forecasting device based on historical cases, characterized by, The device includes: The parameter acquisition module is used to acquire flood prediction data for the watershed to be predicted, wherein the flood prediction data includes watershed type data, flood type data, and watershed environmental data. The case filtering module is used to obtain multiple watershed flood cases from a preset case library based on the watershed type data and the flood type data. The watershed flood cases include historical flood forecast information and historical watershed environmental data. The similarity calculation module is used to determine the case similarity between the historical watershed environmental data and the watershed environmental data of each watershed flood case, obtain the case similarity corresponding to each watershed flood case, and determine the watershed flood cases corresponding to the case similarity higher than the preset similarity threshold as target cases; The weight calculation module is used to determine the similarity weight of each target case based on the case similarity of each target case, and to determine the confidence weight of each target case based on the confidence of the historical flood forecast information of each target case; The forecast correction module is used to correct the historical flood forecast information of each target case based on the similarity weight and confidence weight of the target case, so as to obtain the corrected forecast information of each target case. The result output module is used to determine the flood forecast information of the watershed to be predicted based on the corrected forecast information of each target case.

9. A storage medium having stored thereon a computer program, characterized in that When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 7.

10. A computer comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 7.

Images