A method and device for early flood warning through upstream and downstream linkage

Abstract

The application provides a method and device for early flood warning through upstream and downstream linkage, belongs to the field of flood warning, and is used for solving the problem that flood warning cannot be accurately, timely and reliably performed in the related art.In the method and device, a self-developed algorithm model is used to intelligently analyze key features of water levels of monitoring points in upstream and downstream of a river basin and flood influencing factors, and the key features are gradually and selectively used to analyze flood warning scores, so that the final analysis result of the flood warning scores is more accurate and reliable.The flood warning scores are used to publish the flood warning in advance, which is beneficial to timely, accurate and reliable early publication of the flood warning, and thus the purpose of reducing or even avoiding economic losses and casualties caused by the flood is achieved.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of flood warning, in particular to a method and device for early flood warning through upstream and downstream linkage. BACKGROUND

[0002] Flood is a kind of natural disaster with great destructive power, which occurs frequently all over the world every year. It not only destroys houses, roads, bridges and other infrastructure, resulting in a large amount of property loss, but also seriously threatens people's life safety and causes casualties. Flood disaster also causes heavy damage to agricultural production, floods farmland, destroys crops, affects grain yield, and further affects regional and even national food security and economic stability.

[0003] In order to avoid the serious consequences of economic loss, casualties and other consequences caused by flood, it is necessary to accurately warn the flood disaster. The traditional flood warning system mainly relies on the water level data of a limited number of monitoring points, lacks consideration of the specific circumstances of the river basin and other complex flood influencing factors, and is difficult to accurately, timely and reliably warn the flood. SUMMARY

[0004] The present application provides a method and device for early flood warning through upstream and downstream linkage, which is beneficial to accurately, timely and reliably warning the flood, so as to reduce or even avoid the economic loss and casualties caused by the flood.

[0005] In a first aspect, the present application provides a method for early flood warning through upstream and downstream linkage. The method comprises:

[0006] obtaining water level data and flood influencing factor information of a downstream monitoring point and a plurality of upstream monitoring points in a first preset time period of a river basin;

[0007] substituting the water level data and the flood influencing factor into a pre-constructed water level prediction model to obtain a predicted water level change curve of the downstream monitoring point in a second preset time period after the current time, and analyzing the curve to obtain a predicted key characteristic parameter;

[0008] analyzing the water level data of the downstream monitoring point in the first preset time period to obtain a historical key characteristic parameter;

[0009] determining a basic warning range according to the water level data of the downstream monitoring point at the current time, the pre-obtained warning water level threshold and the historical key characteristic parameter;

[0010] determining a flood warning score in the basic warning range in combination with the flood influencing factor information, the water level data of the upstream monitoring point and the predicted key characteristic parameter, and issuing a flood warning signal when the score exceeds a warning score threshold.

[0011] By adopting the technical scheme, the key features are gradually extracted and the flood warning score is determined by using multi-dimensional data comprehensive analysis, so that the warning result is more accurate and reliable, which is beneficial to timely issuing the warning and reducing the economic loss and casualties caused by the flood.

[0012] Further, the lower limit and the upper limit of the basic warning range are negatively correlated with the difference between the current water level of the downstream monitoring point and the warning water level threshold, and are positively correlated with the cumulative duration, the maximum water level change rate and the cumulative water level change amount in the historical key feature parameters.

[0013] By adopting the technical scheme, the determination of the basic warning range is more in line with the actual water level change rule, laying a foundation for the accurate calculation of the subsequent warning score.

[0014] Further, the determination of the flood warning score includes determining a factor influence coefficient based on flood influencing factors, determining an upstream water level influence coefficient based on upstream water level data, determining a predicted feature influence coefficient based on predicted key feature parameters, and calculating the flood warning score in the basic warning range combined with the above coefficients.

[0015] By adopting the technical scheme, the warning score is calculated by comprehensively considering various influencing factors, which improves the comprehensiveness and accuracy of the score.

[0016] Further, the factor influence coefficient is determined based on the risk coefficients of multi-dimensional flood influencing factors, and each dimension risk coefficient is normalized.

[0017] By adopting the technical scheme, the flood influencing factors in different dimensions can be quantified and have comparability, which improves the reliability of the factor influence coefficient.

[0018] Further, the upstream water level influence coefficient is determined based on the key feature parameters of the tributaries of each upstream monitoring point, and the key feature parameters of the tributaries include the cumulative duration of the water level change rate exceeding a preset value, the maximum water level change rate and the cumulative water level change amount.

[0019] By adopting the technical scheme, the influence of the upstream tributaries on the downstream flood is fully considered, so that the determination of the upstream water level influence coefficient is more scientific.

[0020] Further, the predicted feature influence coefficient is determined based on the predicted key feature parameters, and the predicted key feature parameters include the interval duration of the predicted water level reaching the warning threshold, the cumulative duration of the water level change rate exceeding a preset value and the maximum water level change rate.

[0021] By adopting the technical scheme, the influence coefficient is determined combined with the future water level change trend, which enhances the foresight of the warning.

[0022] Further, the water level prediction model is a neural network model, a decision tree model or a knowledge graph model.

[0023] By adopting the above technical solution, the water level prediction is realized by using a mature model, and the accuracy and stability of the prediction result are ensured.

[0024] Further, the flood warning signal carries warning degree information, and the warning degree information is positively correlated with the flood warning score.

[0025] By adopting the above technical solution, the warning signal more intuitively reflects the flood risk degree, and relevant personnel can take targeted countermeasures.

[0026] In a second aspect, the application provides a device for upstream and downstream linkage to publish flood warning in advance. The device applies any one of the methods described in the first aspect above.

[0027] By adopting the above technical solution, the device can effectively apply the foregoing method to realize accurate and timely flood warning and provide strong support for flood response.

[0028] In summary, the application at least has the following beneficial effects:

[0029] 1. A flood warning method and device for upstream and downstream linkage are provided, realizing accurate, timely and reliable warning;

[0030] 2. The accuracy and reliability of the flood warning score are improved through multi-dimensional data comprehensive analysis;

[0031] 3. Multiple models and influence factors are compatible, and good practicability and expansibility are achieved.

[0032] It should be understood that the content described in the summary section is not intended to limit the key or important features of the embodiments of the application, nor to limit the scope of the application. Other features of the application will become apparent through the following description. BRIEF DESCRIPTION OF DRAWINGS

[0033] The above and other features, advantages and aspects of the embodiments of the application will become more apparent by referring to the following detailed description in conjunction with the accompanying drawings. In the drawings, the same or similar reference numerals indicate the same or similar elements, wherein:

[0034] Figure 1 A flowchart of a method for upstream and downstream linkage to publish flood warning in advance in the embodiments of the application is shown. DETAILED DESCRIPTION

[0035] In order to make the objects, technical solutions and advantages of the embodiments of the present application clearer, the following will clearly and completely describe the technical solutions in the embodiments of the present application with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only some but not all of the embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by a person of ordinary skill in the art without creative work fall within the protection scope of the present application.

[0036] In addition, the term "and / or" in the present application is only used to describe the association relationship of the associated objects, which means that there can be three relationships, for example, A and / or B can mean that A exists alone, A and B exist together, and B exists alone. In addition, the character " / " in the present application generally represents an "or" relationship between the front and rear associated objects.

[0037] The present application provides a method and device for early flood warning by upstream and downstream linkage, which can comprehensively and intelligently analyze flood warning scores based on the long-time series of upstream and downstream water levels of river basins and flood influencing factors of river basins, so as to timely, accurately and reliably publish early flood warning.

[0038] In a first aspect, the embodiments of the present application disclose a method for early flood warning by upstream and downstream linkage. The method can be applied to a server of a flood warning system, and is mainly used for early flood warning when there is a risk of flood in the downstream of a river basin, so as to timely start flood response work and reduce or even avoid economic losses and casualties caused by floods.

[0039] Specifically, the river basin refers to a river catchment area surrounded by a watershed, which is a region with specific hydrological, geographical and ecological characteristics. The river basin mainly includes upstream tributaries and downstream main streams. The upstream tributaries converge to form the downstream main stream, and the downstream main stream generally has only one. Floods usually occur in the downstream main stream, and the main monitoring and warning object of the flood warning system is one or more target points in the downstream main stream. When the target point has a flood, it may cause economic losses and casualties.

[0040] The flood warning system includes a water level monitoring module, a factor monitoring module and a server. The water level monitoring module is configured at the monitoring points of the upstream tributaries and the downstream main stream to collect the water levels of the monitoring points in real time.

[0041] Specifically, water level monitoring modules are arranged at monitoring points of multiple upstream tributaries, and the points of the upstream tributaries where the water level monitoring modules are arranged are upstream monitoring points, and water level data obtained by monitoring is water level data of the upstream tributaries; generally, only one monitoring point of a downstream main stream is arranged for a target point of monitoring and early warning, the monitoring point is at the target point or upstream of the target point, the water level monitoring module is arranged at the monitoring point as a downstream monitoring point, and water level data obtained by monitoring is water level data of the downstream main stream; in the embodiment of the application, the downstream monitoring point is at the target point. The water level monitoring module can be implemented by using a water level gauge, a water gauge, or remote sensing technology, and specific arrangement modes (such as specific selection of arrangement positions and selection of types due to the shape of a river channel and the shape of a cross section) can be implemented by a person skilled in the art, and the implementation modes belong to common knowledge and will not be described here.

[0042] The factor monitoring module is used to monitor factors that may affect floods in a river basin, including climate factors such as rainfall, snowmelt, hurricanes, and typhoons, and topographic and geomorphic features such as basin slope, river channel shape, basin area, soil and rock type, and water system features such as tributary number and distribution, river network density, and human factors such as forest cutting, urbanization process, water conservancy construction, land use change, and even experience of personnel near the target point in dealing with floods, specific conditions of flood control facilities, and the like. In short, all factors that may affect the occurrence and results of floods and can be collected can be considered. The factor monitoring module collects flood influencing factor information, and the flood influencing factor information is multidimensional information. Specific dimensions are not listed and introduced here. Factor information of each dimension can be presented in the form of a space-time matrix, for example, rainfall conditions can be presented in the form of a specific quantified value at a specific time and a specific geographic location (or a local geographic area), and factors that are not sensitive to time (unchanged for a short time or even unchanged for a long time) such as geological conditions can be presented in the form of a space matrix. Key influencing factors in the river basin can be collected in an appropriate form.

[0043] In one example, the factor monitoring module can be configured to include a rain gauge, a geological information input sub-module, a vegetation condition collection sub-module, a reservoir gate state and control plan input sub-module, and a personnel and equipment configuration situation input sub-module, as factor content, such as rainfall conditions, geological conditions, vegetation conditions, reservoir gate control conditions, and personnel and equipment configuration conditions. The rain gauge can be configured in the river basin, and the geological information, vegetation conditions, reservoir gate state and control plan, and personnel configuration situation can be collected and input in a manner based on actual conditions by a person. For example, vegetation conditions can also be analyzed by remote sensing images, and the reservoir gate state can also be achieved by a configured control monitor. The implementation means of each sub-module itself is not the improvement content of the present application, and belongs to the common knowledge of those skilled in the art, and therefore is not enumerated and introduced here.

[0044] For the collected multi-dimensional flood influencing factor information, each dimension can be converted into quantitative dimensional data, and all dimensional data can be normalized to convert the dimensional data of each dimension of the flood influencing factor into a factor risk coefficient between 0 and 1. For each dimension of the flood influencing factor information, a greater factor influencing coefficient indicates that the actual factor situation will promote the occurrence of flood or cause more serious consequences when the flood occurs.

[0045] The server is communicatively connected to the water level monitoring module and the factor monitoring module to realize real-time collection of water level data of the downstream monitoring point and the plurality of upstream monitoring points in the river basin and flood influencing factor information of the river basin, and to store the water level data and the flood influencing factor information in a manner carrying a time stamp, so as to comprehensively, intelligently, reasonably, and selectively analyze the possibility of flood occurrence based on the time-series water level data and the flood influencing factor information, to express in the form of a quantitative flood warning score, and to timely, accurately, and reliably publish the flood warning in advance based on the flood warning score.

[0046] Figure 1 A flowchart of a method for upstream and downstream linkage in advance of publishing a flood warning in the embodiment of the present application is shown.

[0047] Referring to Figure 1 The method specifically includes the following steps:

[0048] S110: Obtain water level data of a downstream monitoring point and a plurality of upstream monitoring points and flood influencing factor information in a first preset time period in a river basin.

[0049] The server collects and stores water level data of the downstream monitoring point and the upstream monitoring points and flood influencing factor information in real time, and determines the water level data and the flood influencing factor information within the first preset time length based on timestamps carried by the water level data and the flood influencing factor information, so as to realize acquisition of the water level data of the downstream monitoring point and the plurality of upstream monitoring points and the flood influencing factor information within the first preset time length in the river basin.

[0050] Of course, the acquired water level data and flood influencing factor information need to be preprocessed, specifically including: smoothing the water level data, for example, removing instantaneous fluctuation noise by using a sliding average method (the window size is set to 5-10 minutes, which can be adjusted according to the monitoring frequency), for example, smoothing the water level data of the upstream monitoring point at a certain moment , the processed data is is the window length); normalizing and standardizing the flood influencing factor information, for example, for continuous factors such as rainfall and snowmelt, min-max normalization is used to interval (formula: , wherein is the original value, is the historical extreme value); for discrete factors such as basin slope, the risk coefficient is converted to the 0-1 interval by expert scoring method; missing value processing, for example, if the data of a monitoring point is missing for minutes, linear interpolation is used to complete the missing data; if the missing time length is minutes, the standby monitoring point data (if any) is used or marked as an abnormal period, and is temporarily not included in the current analysis.

[0051] S120: substituting the water level data and the flood influencing factor into a pre-constructed water level prediction model to obtain a predicted water level change curve of the downstream monitoring point within a second preset time length after the current moment, and analyzing the curve to obtain a predicted key characteristic parameter.

[0052] In the method of this step, the water level prediction model is a neural network model, a decision tree or a knowledge graph model; by substituting the water level data and the flood influencing factor into the model, a predicted water level change curve of the downstream monitoring point within a second preset time length after the current moment is obtained, and then the curve is analyzed to obtain a predicted key characteristic parameter.

[0053] The water level prediction model can also be other self-designed interpretable machine learning models, or other existing black box models, or more intelligent models combined with large models, as long as they can realize the time series water level prediction of the downstream monitoring point based on the water level data of the downstream monitoring point and the upstream monitoring point and the flood influencing factors of the river basin to obtain the predicted water level change curve. For interpretable machine learning models, the structure of the machine learning model can be designed and adjusted based on the professional knowledge of hydrologists and the historical data of the river basin, and the historical data is used to train the parameters of the machine learning model. For the black box model, the historical data of the river basin or the simulation data of the professional knowledge of hydrologists can be used for training.

[0054] In one specific example, the water level prediction model is a decision tree model, the experience of hydrologists and the historical data of the river basin are extracted and classified to construct decision rules, and then the root node, intermediate node, leaf node, etc. are determined, so that the water level prediction model meets the professional knowledge and experience of hydrologists and the historical data of the river basin.

[0055] In another specific example, the water level prediction model is a neural network model, the historical data of the river basin is used to train the neural network model, and the water level data of the downstream monitoring point and the downstream monitoring point within the first preset time length before the specified time and the influencing factor information of the river basin are used as the input data for training, and the water level data of the downstream monitoring point within the second preset time length after the specified time is used as the output for training. The training of the parameters in the neural network model is realized.

[0056] Other specific implementation examples of the water level prediction model are not listed and introduced one by one, and those skilled in the art can configure the water level prediction model according to common knowledge and actual needs.

[0057] After obtaining the predicted water level change curve, the predicted key feature parameters can be extracted and analyzed in the predicted water level change curve. In the method of this step, the predicted key feature parameters include one or more of the interval time between the future time when the predicted water level reaches the pre-obtained warning water level threshold and the current time, the cumulative time when the water level change rate exceeds the preset change rate, and the maximum water level change rate. For the determination of the interval time, the time difference between the future time when the predicted water level reaches the warning water level threshold and the current time can be calculated; for the determination of the cumulative time, the tangent slope (derivative of the continuous predicted water level change curve) of each point of the predicted water level change curve is taken as the water level change rate, and the cumulative time (time integral) when the water level change rate exceeds the preset change rate is calculated; for the determination of the maximum water level change rate, the maximum value in the water level change rate can be taken.

[0058] In one specific example, the predicted key characteristic parameters include the interval length between the future time when the water level reaches the pre-obtained early warning water level threshold and the current time, the accumulated length of the water level change rate exceeding the preset change rate, and the maximum water level change rate.

[0059] Of course, the preset key characteristic parameters can also include only one or two of the aforementioned three parameters, or other unlisted parameters, as long as they can be extracted in the predicted water level change curve.

[0060] The pre-constructed water level prediction model needs to be trained and optimized by historical data. The specific process includes: the training data source needs to use the measured water level data of the river basin in the past 5-10 years (including the wet season and the dry season) and the flood influencing factor information in the corresponding period, and the training set and the validation set are divided in the ratio of 7:3; the model parameter initialization should take the neural network model as an example, the input layer dimension is (the number of upstream monitoring points + the number of downstream monitoring points + the dimension of flood influencing factors), 2-3 layers of hidden layers are set (the number of neurons in each layer is 1.5-2 times of the input layer), and the output layer is the water level sequence of the downstream monitoring point in the future second preset time length; the model optimization should take the mean square error (MSE) of the predicted water level and the actual water level as the loss function, use the Adam optimizer (the learning rate is initially set to 0.001, and is attenuated by 10% every 100 rounds) for iterative training, until the validation set MSE tends to be stable (fluctuation ≤5%); the model update should pay attention to fine-tuning the model based on the newly measured data every quarter to ensure the adaptation to the long-term changes of the basin hydrological characteristics (such as river channel siltation, change of vegetation coverage, expansion of impervious area caused by urbanization process, new construction or reconstruction of water conservancy facilities, change of land use type, and change of precipitation pattern caused by climate change, etc.).

[0061] S130: Analyze the water level data of the downstream monitoring point in the first preset time length to obtain the historical key characteristic parameters.

[0062] In the method of this step, the historical key characteristic parameters include one or more of the accumulated length of the water level change rate exceeding the preset change rate, the maximum water level change rate, and the accumulated water level change amount. For specific disclosure of the historical key characteristic parameters, it can be specifically analogous to the specific disclosure of the predicted key characteristic parameters, which is not repeated here.

[0063] In the embodiments of the present application, the historical key characteristic parameters include the accumulated length of the water level change rate exceeding the preset change rate, the maximum water level change rate, and the accumulated water level change amount.

[0064] S140: Determine the basic early warning range according to the water level data of the downstream monitoring point at the current time, the pre-obtained early warning water level threshold, and the historical key characteristic parameters.

[0065] In the method of the present step, the determining the basic warning range comprises: the lower limit and the upper limit of the basic warning range are negatively correlated with the difference between the current water level of the downstream monitoring point and the warning water level threshold, and are positively correlated with the accumulated time length, the maximum water level change rate and the accumulated water level change amount in the historical key characteristic parameters.

[0066] In one example, let the water level data of the downstream monitoring point at the current time be , the pre-obtained warning water level threshold be , the accumulated time length, the maximum water level change rate and the accumulated water level change amount in the historical key characteristic parameters be , , , the range lower limit and the range upper limit of the basic warning range be , , then

[0067]

[0068]

[0069]

[0070]

[0071] In the formula, , , ,..., are preset values and , , , , respectively represent a historical time length calculation weight, a historical change rate calculation weight and a historical change amount calculation weight which are greater than zero and are preset relative to the accumulated time length, the maximum water level change rate and the accumulated water level change amount in the historical key characteristic parameters, , , ,..., are preset values and . In the embodiments of the present application, , so as to realize quantification of the emergency degree of the flood warning to an emergency degree score between 0 and 100, that is, the score represents the emergency degree. Further, the range lower limit and the range upper limit of the basic warning range are determined in combination with the historical key characteristic parameters and the emergency degree score, so that the basic warning range is a sub-range between 0 and 100, and further so that the final flood warning score is between 0 and 100.

[0072] When the content of the historical key characteristic parameters changes, the model can also take other forms, for example or or only adjust the structure to and so on; the emergency degree score can also be in other forms, such as and scaling the result to the target range.

[0073] S150: In combination with the flood influencing factor information, the water level data of the upstream monitoring point and the predicted key characteristic parameters, the flood warning score is determined within the basic warning range, and when the score exceeds the warning score threshold, the flood warning signal is issued.

[0074] In the method of this step, the determination of the flood warning score includes: determining the factor influence coefficient based on the flood influencing factors, determining the upstream water level influence coefficient based on the water level data of the upstream monitoring point, determining the predicted characteristic influence coefficient based on the predicted key characteristic parameters, and combining the above coefficients to calculate the flood warning score within the basic warning range; the factor influence coefficient is determined based on the risk coefficient of the multi-dimensional flood influencing factor, and each dimension risk coefficient is normalized; the upstream water level influence coefficient is determined based on the key characteristic parameters of the tributaries of each upstream monitoring point, and the key characteristic parameters of the tributaries include the cumulative duration of the water level change rate exceeding the preset value, the maximum water level change rate and the cumulative water level change amount; the predicted characteristic influence coefficient is determined based on the predicted key characteristic parameters, and the predicted key characteristic parameters include the interval duration of the predicted water level reaching the warning threshold, the cumulative duration of the water level change rate exceeding the preset value and the maximum water level change rate; the flood warning signal carries the warning degree information, and the warning degree information is positively correlated with the flood warning score.

[0075] In one example, the method of this step specifically includes: setting the lower limit and upper limit of the range of the basic warning range as , , and the flood warning score is ,

[0076]

[0077] In the formula, is the factor influence coefficient determined based on the flood influencing factors, is the upstream water level influence coefficient determined based on the water level data of the upstream monitoring point, is the predicted characteristic influence coefficient determined based on the predicted key characteristic parameters, , , , , , The factor coefficient weight, the upstream influence weight, and the prediction characteristic weight are respectively a preset factor coefficient weight greater than 0, an upstream influence weight, and a prediction characteristic weight, = 1. Based on this, the flood warning score can be determined within the basic warning range.

[0078] The method for determining the factor influence coefficient based on the flood influencing factor includes: assuming that the flood influencing factor has The factor risk coefficient of the ith dimension is , ,

[0079]

[0080] In the formula, is a positive integer and , is a function of sorting from large to small and taking the sum of the first Based on this method, the usability of the factor influence coefficient can be effectively improved, that is, when the dimension of the flood influencing factor is increased or decreased, the calculation model of the factor influence coefficient does not need to be adjusted.

[0081] The method for determining the upstream water level influence coefficient based on the water level data of the upstream monitoring point includes: analyzing the branch key characteristic parameters of the water level data of each upstream monitoring point within a first preset time length, and the branch key characteristic parameters include the cumulative duration of the water level change rate exceeding the corresponding branch preset change rate, the maximum water level change rate, and the cumulative water level change amount. Here, the analysis of the branch key characteristic parameters of the water level data of each upstream monitoring point can refer to the foregoing disclosure of the analysis of the prediction key characteristic parameters and the historical key characteristic parameters, and no repeated disclosure is made.

[0082] The upstream water level influence coefficient is determined in combination with the branch key characteristic parameters of each upstream monitoring point. Assuming that there are upstream monitoring points, the cumulative duration, the maximum water level change rate, and the cumulative water level change amount in the branch key characteristic parameters of the ith upstream monitoring point are respectively , , ,

[0083]

[0084]

[0085] In the formula, , , ​​are respectively a branch length weight, a branch rate weight and a branch amount weight which are greater than zero and are preset for the cumulative length, the maximum water level change rate and the cumulative water level change amount in the relative branch key characteristic parameter, to be sorted from large to small and take the sum of the first . Similarly, this part of the algorithm structure has good scalability, whether it is the obtained from the branch key characteristic parameter of a single upstream monitoring point or the number of upstream branch monitoring points can be flexibly adjusted.

[0086] The method for determining the prediction characteristic influence coefficient based on the prediction key characteristic parameter comprises: setting the interval length, the cumulative length and the maximum water level change rate in the prediction key characteristic parameter as , , , then

[0087]

[0088] In the formula, , are respectively a prediction length weight and a prediction rate weight which are greater than zero and are preset for the cumulative length and the maximum water level change rate in the relative prediction key characteristic parameter.

[0089] In other examples, the specific algorithm model for determining the flood warning score in the basic warning range in combination with the flood influence factor information, the water level data of the upstream monitoring point and the prediction key characteristic parameter can also be in other forms, for example, only one or two of the flood influence factor information, the water level data of the upstream monitoring point and the prediction key characteristic parameter are considered, which can make or , when the flood influence factor information, the water level data of the upstream monitoring point and the prediction key characteristic parameter are considered, the calculation model of the flood warning score can also be adjusted to make , only the correlation between the flood influence factor information, the water level data of the upstream monitoring point and the prediction key characteristic parameter and the flood warning score needs to be followed, and the same model structure transformation method is also applicable to , , , the calculation of itself, the content considered in the adjustment, such as the content of the branch key characteristic parameter, the content of the prediction key characteristic parameter, etc., so that the model structure has multiple forms, which are not listed and introduced here.

[0090] Further, when a subsequent abnormal scene occurs, the flood warning score calculation logic can be adjusted:

[0091] In scenarios where upstream monitoring point data is interrupted, if data from a single upstream monitoring point is interrupted, the average value of historical data from that monitoring point for the same period will be temporarily used to replace the key characteristic parameters of its tributary; if... Data interruption at upstream monitoring points will affect the weighting of the upstream water level influence coefficient. Reduced to the original weight And it preferentially relies on the influence coefficient of the predicted features;

[0092] Predicting scenarios of sudden water level changes, such as the magnitude of the change in water level at a certain moment compared to the previous moment in the predicted water level change curve. (Exceeding historical extremes), then the maximum water level change rate weight in the predicted characteristic influence coefficient will be ( ) Upward adjustment And trigger secondary verification (re-substitute into the model for calculation);

[0093] In scenarios where warning scores fluctuate, if the flood warning score fluctuation range is calculated for three consecutive times... If the score is 100 points (for example), the first preset time period will be extended (e.g., from 24 hours to 48 hours), and the historical key feature parameters will be re-analyzed before the score is determined.

[0094] The flood warning score is determined in real time. After the flood warning score is determined, it can be compared with the pre-configured warning score threshold in the server in real time. When the flood warning score exceeds the warning score threshold, a flood warning signal is issued to relevant personnel. For example, the flood warning signal can be broadcast using loudspeakers configured at downstream monitoring points or sent to the smart terminals of relevant personnel. The flood warning signal can carry the flood warning score, or the flood warning score (or the excess value of the flood warning score exceeding the warning score threshold, or the ratio of the excess value to the warning score threshold, etc.) can be converted into semi-quantitative warning level information, such as a level one to five warning, so that relevant personnel can respond to the flood warning with different strategies.

[0095] Considering that one or more factors can be selected in the methodology model, such as flood influencing factors and key prediction characteristic parameters, it is advisable to pre-screen these factors when conducting flood prediction tasks for river basins. To optimize the factor screening logic and balance prediction accuracy with computational speed, a multi-objective optimization mechanism can be introduced before calculating the factor influence coefficients. Specifically, causal reasoning is first used to quantify the standardized average treatment effect of each factor on downstream water levels. ,in As factors Average treatment effect on downstream water level changes The higher the value, the greater the causal importance of the factor.

[0096] For a single factor Setting experience threshold to influence prediction accuracy When , its contribution to the sum of causal importance is , which can significantly improve prediction accuracy; when , , it is considered as a redundant factor, and its addition has negligible effect on accuracy improvement, and may even introduce noise.

[0097] In terms of calculation speed, the time consumed by single-round prediction is , and the addition of a single factor will increase the time consumption by , which is composed of preprocessing time and model input time . Continuous factors (such as rainfall) need to be cleaned of outliers, and steps such as moving average denoising, about about ; discrete factors (such as watershed slope) need to be encoded, about about ; spatiotemporal coupling factors (such as spatiotemporal distribution of rainfall) need to be spatiotemporally interpolated, up to , about . When the number of factors exceeds the empirical value , a correction factor of 1.5 is needed to reflect the nonlinear increase in time consumption caused by the explosion of dimensions.

[0098] Based on the above influences, a double-objective optimization model is constructed: maximize the sum of causal importance of the factor subset , which represents prediction accuracy, and minimize the number of factors , which represents calculation speed. The Pareto optimal solution is solved by NSGA-II algorithm. In practical applications, the optimal subset can be selected according to the scene: in emergency scenarios, the subset with smaller and higher is preferred, while in high-precision scenarios, the subset with larger is preferred, so as to ensure the effectiveness of the warning while meeting the real-time requirements.

[0099] This method is an interpretable machine learning model designed independently, and its model structure is based on the professional knowledge and experience of flood warning and hydrological industry experts, with good scalability and adjustability and compatibility. The preset values in it, such as , , , , , , , etc. and calculating weights, for example , , , , , , , , , , , etc. support human adjustment or data-based training, enabling the method to link upstream and downstream water level data, intelligently and comprehensively analyze the possibility of flood occurrence in the river basin in combination with multiple factors of the river basin, and express the possibility in the form of a quantitative flood warning score, and then publish the flood warning in advance in combination with the warning score threshold, which is conducive to timely, accurate, and reliable publication of the flood warning in advance, so as to protect the lives and property safety of personnel in the flood-affected area.

[0100] Of course, the core parameters can also be dynamically adjusted according to different basin characteristics. For example, in the case of a mountainous basin (slope ), the second preset time is shortened to 6 hours (12 hours for a plain basin) due to the fast confluence speed, and the weight of the upstream water level influence coefficient ( ) is increased to 0.4 (0.3 for a plain basin); in the case of a city basin (the proportion of impervious area ), the warning score threshold is lowered to 60 points (70 points for a rural basin) due to the fast flood response, and the weight of the 'urbanization process' dimension in the factor influence coefficient is increased ; in the case of a seasonal river, the lower limit of the basic warning range ( ) is lowered to of the original lower limit during the dry season (runoff < 10% of the annual average), to reduce false positives.

[0101] To facilitate understanding of the specific calculation process of the above method, a specific scenario is used for example illustration. The scenario takes a certain plain basin as the monitoring object, wherein the first preset time is set to 24 hours for analyzing historical water level data and flood influencing factor information; the second preset time is set to 12 hours for predicting future water level changes at the downstream monitoring point; the warning water level threshold is 10.0 m, and the warning score threshold is 60 points (100 points in total); there are 3 upstream monitoring points (denoted as ), and the flood influencing factors include rainfall, basin slope, vegetation coverage, and reservoir regulation (denoted as ).

[0102] The specific input data are as follows: the current water level of the downstream monitoring point 8.5m; among the historical key characteristic parameters, the accumulated length of time when the water level change rate exceeds the preset value Maximum water level change rate per hour Accumulated water level change ; among the upstream monitoring point data, Accumulated length of time Maximum water level change rate per hour Accumulated water level change Accumulated length of time Maximum water level change rate per hour Accumulated water level change Accumulated length of time Maximum water level change rate per hour Accumulated water level change ; among the predicted key characteristic parameters, the interval length of time when the predicted water level reaches the warning threshold Accumulated length of time when the water level change rate exceeds the preset value Maximum water level change rate per hour After the risk coefficients of the flood influencing factors are normalized, (rainfall), (basin slope) (vegetation coverage) (reservoir regulation).

[0103] The weight parameters are set as follows: among the historical key characteristic parameters, the accumulated length of time weight Maximum water level change rate weight Accumulated water level change weight ; among the influence coefficients, the factor influence coefficient weight Upstream water level influence coefficient weight Predicted characteristic influence coefficient weight ; among the tributary key characteristic parameters, the accumulated length of time weight Maximum water level change rate weight Accumulated water level change weight ; among the predicted characteristic parameters, the accumulated length of time weight Maximum water level change rate weight .

[0104] Based on the above parameters, the specific calculation process is as follows: first, calculate the basic warning range, the difference between the current water level of the downstream monitoring point and the warning water level threshold is , corresponding to the preset function ; the historical characteristic parameter combination value , ; take the preset coefficient , then the upper limit of the basic warning range is , and the lower limit is i.e. the basic early warning range is [21.6, 37.6].

[0105] Then the flood early warning score is calculated, and the factor influence coefficient Based on the risk coefficient calculation of the four dimensions, take Then ; the upstream water level influence coefficient The key characteristic parameter combination value of each tributary needs to be calculated first, wherein , , Take Then ; the predicted characteristic influence coefficient .

[0106] The final flood early warning score = points. The score does not exceed the early warning score threshold of 60 points, so no flood early warning signal is issued. It should be noted that the preset parameters (such as , various weight values, etc.) in the above example can be adjusted according to the actual hydrological characteristics, historical data and expert experience of the basin to adapt to the early warning needs of different scenarios.

[0107] It should be noted that for the foregoing method embodiments, in order to simply describe, they are all expressed as a series of action combinations, but those skilled in the art should know that the present application is not limited by the order of the described actions, because according to the embodiments of the present application, certain steps can be performed in other orders or simultaneously. Secondly, those skilled in the art should know that the embodiments described in the specification are all optional embodiments, and the actions and modules involved are not necessarily required by the present application.

[0108] In a second aspect, the embodiments of the present application disclose a device for early flood warning through upstream and downstream linkage. The device specifically applies the method disclosed in the first aspect of the present application. The device can be configured as a server in a flood early warning system, or contained in a server in a flood early warning system.

[0109] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working process of the described device can refer to the corresponding process in the foregoing method embodiments, which will not be repeated here.

[0110] In summary, the present application at least contains the following beneficial effects:

[0111] 1. A method and device for providing an upstream and downstream linkage early flood warning are provided, which is based on a self-developed algorithm model for comprehensively, intelligently and gradually analyzing water levels and flood influencing factors of monitoring points in upstream and downstream river basins with focus, so as to accurately determine a flood warning score, and is beneficial to timely, accurate and reliable early flood warning effect;

[0112] 2. The basic warning range is determined by using the time sequence water level and the current water level of the downstream monitoring point and the warning water level threshold, and the flood warning score is determined in the basic warning range by using the predicted water level of the downstream monitoring point, the time sequence water level of the upstream monitoring point and the flood influencing factors, which is beneficial to further improve the accuracy and reliability of the flood warning score;

[0113] 3. The algorithm model of multi-dimensional data comprehensive intelligent analysis is adopted, the model structure of the algorithm model is configured, the final flood warning score result is ensured to be in a preset value range, which is convenient for quantitative evaluation by the person skilled in the art, and the interpretable machine learning model is also beneficial to the person skilled in the art to configure parameters and adjust the model structure.

[0114] The above description is only the preferred embodiment of the present application and the explanation of the applied technical principles. It should be understood by those skilled in the art that the disclosed range in the present application is not limited to the technical solutions formed by the specific combination of the above technical features, and also covers other technical solutions formed by any combination of the above technical features or equivalent features without departing from the disclosed concept. For example, the above features are replaced with the technical features disclosed in the present application (but not limited to) having similar functions to form technical solutions.

Claims

1. A method for coordinated flood warning issuance between upstream and downstream areas, characterized in that, include: Acquire water level data and flood influencing factors from downstream monitoring points and multiple upstream monitoring points within the first preset time period of the river basin; Substitute the water level data and flood influencing factors into the pre-constructed water level prediction model to obtain the predicted water level change curve within the second preset time after the current moment of the downstream monitoring point, and analyze the curve to obtain the key predictive feature parameters. Historical key characteristic parameters were obtained by analyzing water level data from downstream monitoring points within the first preset time period. The basic warning range is determined based on the current water level data of downstream monitoring points, the pre-acquired warning water level threshold, and historical key characteristic parameters. By combining information on flood influencing factors, water level data from upstream monitoring points, and key predictive parameters, a flood warning score is determined within the basic warning range. A flood warning signal is issued when the score exceeds the warning score threshold. The lower and upper limits of the basic warning range are negatively correlated with the difference between the current water level at the downstream monitoring point and the warning water level threshold, and positively correlated with the cumulative duration, maximum water level change rate, and cumulative water level change in historical key characteristic parameters. The determination of the flood warning score includes: determining the factor influence coefficient based on flood influencing factors, determining the upstream water level influence coefficient based on upstream water level data, determining the prediction characteristic influence coefficient based on predicted key characteristic parameters, and calculating the flood warning score within the basic warning range by combining the above coefficients. The factor influence coefficient is determined based on the risk coefficient of multi-dimensional flood influencing factors, and the risk coefficient of each dimension is normalized. The upstream water level influence coefficient is determined based on the key characteristic parameters of the tributaries at each upstream monitoring point. The key characteristic parameters of the tributaries include the cumulative duration of the water level change rate exceeding the preset value, the maximum water level change rate, and the cumulative water level change. The prediction characteristic influence coefficient is determined based on the prediction key characteristic parameters, which include the interval duration of the predicted water level reaching the warning threshold, the cumulative duration of the water level change rate exceeding the preset value, and the maximum water level change rate. Let the current water level data at the downstream monitoring point be... The pre-acquired warning water level threshold is The cumulative duration, maximum water level change rate, and cumulative water level change in the historical key characteristic parameters are as follows: , , The lower and upper limits of the basic early warning range are respectively , ,but In the formula, , , ... All are preset values ​​and , , , , These are the weights for calculating the cumulative duration, maximum water level change rate, and cumulative water level change amount, respectively, based on preset historical duration values ​​greater than zero, historical change rate calculation weights, and historical change amount calculation weights among the relative historical key characteristic parameters. Let the flood warning score be... ,but In the formula, These are the factor influence coefficients determined based on flood influencing factors. The upstream water level influence coefficient is determined based on water level data from upstream monitoring points. The predicted feature influence coefficient is determined based on the predicted key feature parameters. , , , , , These are the relative factor influence coefficient, upstream water level influence coefficient, and prediction feature influence coefficient, respectively, along with the preset weights of factor coefficients greater than zero, upstream influence weight, and prediction feature weight. + + =1; Among them, the methods for determining the factor influence coefficients based on flood influencing factors include: assuming there are flood influencing factors... There are 3 dimensions, and the risk coefficient of the factor in the i-th dimension is 1. , ,but In the formula, are positive integers and , To be Sort by size from largest to smallest and take the first few. A function of the sum of ; The upstream water level influence coefficient is determined by combining the key characteristic parameters of the tributary at each upstream monitoring point. Assume there are upstream monitoring points... Among the key characteristic parameters of the tributary at the i-th upstream monitoring point, the cumulative duration, maximum water level change rate, and cumulative water level change are respectively... , , ,but In the formula, , , These are the preset weights for tributary duration, tributary change rate, and tributary change amount that are greater than zero, respectively, among the key characteristic parameters of relative tributaries: cumulative duration, maximum water level change rate, and cumulative water level change. To be Sort by size from largest to smallest and take the first few. A function of the sum of ; Methods for determining the influence coefficient of prediction features based on key prediction feature parameters include: Let the interval duration, cumulative duration, and maximum water level change rate among the key prediction feature parameters be respectively... , , ,but In the formula, , These are the preset weights of the cumulative duration and the maximum water level change rate, which are respectively greater than zero, for the key characteristic parameters of relative prediction.

2. The method according to claim 1, characterized in that, The water level prediction model is a neural network model, a decision tree model, or a knowledge graph model.

3. The method according to claim 1, characterized in that, The flood warning signal carries warning level information, which is positively correlated with the flood warning score.

4. A device for early flood warning issuance through upstream and downstream linkage, characterized in that, The method described in any one of claims 1-3 is applied.

Images