A flood forecasting data determination method, device and equipment and storage medium

Abstract

The application discloses a flood forecasting data determination method, device and equipment and a storage medium. The technical scheme of the application embeds meteorological and hydrological physical rules in a preset flood forecasting large model, integrates historical multi-source flood measurement data corresponding to a historical period and historical initial meteorological forecasting data output by a preset meteorological model, constructs a training sample, trains the preset flood forecasting large model, obtains a target flood forecasting large model, takes target initial meteorological forecasting data of a target period as a basis, corrects the target initial meteorological forecasting data by using the target flood forecasting large model, and checks the meteorological and hydrological physical rules, so as to obtain target flood forecasting data of the target period. The technical scheme of the application improves the accuracy and physical rationality of the flood forecasting data.

Description

Technical Field

[0001] This invention relates to the field of urban flood forecasting technology, and in particular to a method, apparatus, equipment and storage medium for determining flood forecast data. Background Technology

[0002] Existing flood forecasting methods in Hongcheng City lack meteorological, hydrological, and physical constraints, easily producing forecast results that deviate from reality. Furthermore, they rely solely on single observation data without deep integration with pre-set meteorological model data, resulting in poor error correction. They only perform shallow post-processing on the output of the pre-set meteorological model without collaborative reasoning. After model fusion, they are applied statically without iterative optimization, leading to easy degradation of accuracy. None of the existing solutions can meet the core requirements of minute-level, high-precision, and highly reliable urban flood forecasting.

[0003] Therefore, there is an urgent need for a dedicated urban flood forecasting method that is tailored to urban scenarios and deeply integrates numerical models and large-scale models. Summary of the Invention

[0004] This invention provides a method, apparatus, equipment, and storage medium for determining flood forecast data, in order to improve the accuracy and physical rationality of flood forecast data.

[0005] In a first aspect, embodiments of the present invention provide a method for determining flood forecast data, the method comprising: The system determines at least two historical periods corresponding to the target area, including historical multi-source flood measured data and historical initial meteorological forecast data output by a preset meteorological model, target initial meteorological forecast data output by a preset meteorological model for the target period, and meteorological, hydrological, and physical rules adapted to urban flood forecasting scenarios. The meteorological, hydrological, and physical rules include at least a no-flood constraint rule and a rainfall intensity-inundation level matching constraint rule. The meteorological and hydrological physical rules are embedded into a preset flood forecasting model, and the preset flood forecasting model is trained based on the historical initial meteorological forecast data and historical multi-source flood measured data to obtain the target flood forecasting model. The initial meteorological forecast data of the target area is input into the target flood forecast model, and the target flood forecast data of the target area for the target period is output.

[0006] Secondly, embodiments of the present invention also provide a flood forecast data determination device, the device comprising: The data determination module is used to determine the historical multi-source flood measured data and the historical initial meteorological forecast data output by the preset meteorological model for at least two historical time periods in the target area, the target initial meteorological forecast data output by the preset meteorological model for the target time period, and the meteorological and hydrological physical rules adapted to the urban flood forecast scenario; the meteorological and hydrological physical rules include at least the no-flood constraint rule and the rainfall intensity and inundation water level matching constraint rule; The target model determination module is used to embed the meteorological and hydrological physical rules into the preset flood forecasting model, and to train the preset flood forecasting model based on the historical initial meteorological forecast data and historical multi-source flood measured data to obtain the target flood forecasting model. The flood forecast data determination module is used to input the initial meteorological forecast data of the target into the target flood forecast big model and output the target flood forecast data of the target area for the target period.

[0007] Thirdly, embodiments of the present invention also provide an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the flood forecast data determination method as described in any of the embodiments of the present invention.

[0008] Fourthly, embodiments of the present invention also provide a storage medium for storing computer-executable instructions, which, when executed by a computer processor, are used to perform the flood forecast data determination method as described in any of the embodiments of the present invention.

[0009] The technical solution of this invention embeds meteorological and hydrological physical rules into a preset flood forecasting model, integrates historical multi-source flood measurement data corresponding to historical time periods with historical initial meteorological forecast data output by the preset meteorological model, constructs training samples to train the preset flood forecasting model, obtains a target flood forecasting model, and uses the target initial meteorological forecast data for the target time period as a basis to perform data correction and meteorological and hydrological physical rule verification on the target initial meteorological forecast data to obtain the target flood forecast data for the target time period, thereby improving the accuracy and meteorological and hydrological physical rationality of the flood forecast data.

[0010] It should be understood that the description in this section is not intended to identify key or essential features of the embodiments of the present invention, nor is it intended to limit the scope of the invention. Other features of the invention will become readily apparent from the following description. Attached Figure Description

[0011] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0012] Figure 1 This is a flowchart of a method for determining flood forecast data provided in Embodiment 1 of the present invention; Figure 2 This is a schematic diagram of the structure of a flood forecast data determination device provided in Embodiment 2 of the present invention; Figure 3 This is a schematic diagram of the structure of an electronic device that implements the flood forecast data determination method of this invention. Detailed Implementation

[0013] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention.

[0014] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this invention are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of the invention described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover a non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0015] Example 1 Figure 1 This is a flowchart illustrating a method for determining flood forecast data according to Embodiment 1 of the present invention. This embodiment is applicable to situations requiring the determination of flood forecast data. The method can be executed by a flood forecast data determination device, which can be implemented in hardware and / or software. This device can be configured in any electronic device with network communication and computing capabilities. Figure 1 As shown, the method includes: S110. Determine at least two historical periods corresponding to the target area, including historical multi-source flood measured data and historical initial meteorological forecast data output by the preset meteorological model, target initial meteorological forecast data output by the preset meteorological model for the target period, and meteorological and hydrological physical rules adapted to urban flood forecast scenarios; the meteorological and hydrological physical rules include at least the no-precipitation-no-flood constraint rule and the rainfall intensity and inundation level matching constraint rule.

[0016] In this embodiment, the target area can be a specified urban spatial range, such as the main urban area or central urban area of ​​a prefecture-level city. A unified grid resolution is used, and radar, meteorological station, hydrological station, and meteorological model forecast fields are all cropped to match this area to ensure a unified data spatial range. Historical time periods are continuous time intervals that occurred in the past, have been completed, and have complete observation archives. At least two segments are selected, covering various flood weather processes such as no rain, light rain, heavy rain, and torrential rain, including historical heavy rain periods during the flood season and normal periods outside the flood season. Each historical time period includes a complete start time, forecast lead time, and synchronous measured records, enabling a one-to-one spatiotemporal pairing of meteorological forecasts and measured data.

[0017] Historical multi-source flood measurement data refers to the true observation data collected in real time by on-site monitoring equipment and archived afterward for various historical periods in the target area. The multi-source flood measurement data includes at least radar measurement data, meteorological station measurement data, and hydrological station measurement water level data. Radar measurement data includes radar measured instantaneous rainfall rate, measured rainfall for each period, and average rainfall over the basin area. Meteorological station measurement data includes meteorological station measured air temperature, meteorological station measured relative humidity, and meteorological station measured surface air pressure. Hydrological station measurement water level data includes measured river water depth, measured river water level rise, measured river water level change rate, measured instantaneous flow rate at river cross-section, and measured surface inundation depth.

[0018] Historical initial meteorological forecast data consists of gridded time-series simulation forecast data archived after the regional numerical pre-set meteorological model has been running continuously for each historical period of the target area. The initial meteorological forecast data includes rainfall forecast data and wind field forecast data output by the pre-set meteorological model. The rainfall forecast data includes instantaneous rainfall rate, cumulative rainfall for each period, rainfall type classification, and rainfall duration. The wind field forecast data includes horizontal wind speed at a pre-set surface height, horizontal wind azimuth, and instantaneous wind speed. The initial meteorological forecast data also includes the start time, forecast lead time, and grid coordinates. Target initial meteorological forecast data is a gridded meteorological forecast field for the future target period, generated in real-time by the pre-set meteorological model. The pre-set meteorological model is a pre-selected computer numerical simulation system built based on atmospheric dynamics and thermodynamic physical equations. In this embodiment, the meteorological model can be the GRAPES-MESO meteorological model from the meteorological bureau, or the internationally used WRF-Hydro urban hydrological module meteorological model.

[0019] The meteorological and hydrological physical rules are a priori constraints derived from objective hydrological laws governing atmospheric dynamics and urban hydrological runoff generation. These objective hydrological laws include: when there is no effective rainfall replenishment, there is no surface runoff or water accumulation, and the grid inundation depth should approach zero; objective hydrological laws also include: within the same grid and time period, rainfall intensity and surface water level rise should be positively correlated; the stronger the rainfall, the greater the rise in water level / inundation depth should be; significant water accumulation during light rain and almost no water level rise during heavy rainfall both violate physical laws. The meteorological and hydrological physical rules in this embodiment include at least the no-rainfall-no-flood constraint rule and the rainfall intensity-inundation-level matching constraint rule.

[0020] S120. Embed the meteorological and hydrological physical rules into the preset flood forecasting model, and train the preset flood forecasting model based on the historical initial meteorological forecast data and historical multi-source flood measured data to obtain the target flood forecasting model.

[0021] In this embodiment, the preset flood forecasting model is a pre-set neural network model used to generate flood forecasting data. The target flood forecasting model is a neural network model trained from the preset flood forecasting model.

[0022] In this embodiment, differentiable meteorological and hydrological physical rules are first encapsulated into an independent physical constraint layer and embedded into a pre-built preset flood forecasting model. Then, the historical initial meteorological forecast data output by the preset meteorological model and historical multi-source measured flood data are integrated. After spatiotemporal standardization pairing, a fusion training sample is constructed. The modified preset flood forecasting model can be iteratively trained using a dual-objective mechanism of data fitting accuracy and physical rule compliance rate. After the model loss represented by data fitting accuracy and physical rule compliance rate converges, a target flood forecasting model that can be used for refined urban flood forecasting is obtained.

[0023] Optionally, before embedding the meteorological and hydrological physical rules into a preset flood forecasting model and training the preset flood forecasting model based on the historical initial meteorological forecast data and historical multi-source flood measurement data to obtain the target flood forecasting model, the following steps are included: The historical multi-source flood measured data and historical initial meteorological forecast data are preprocessed, including cleaning and denoising, missing value filling, spatiotemporal coordinate unification, and time dimension alignment.

[0024] In this embodiment, noise removal and cleaning are performed on historical weather forecast grid data, radar, meteorological station, and hydrological station multi-source measured data to verify the range and temporal abrupt change, eliminating abnormal noise data generated by various sensors and numerical simulations; missing value filling addresses the issues of missing grid measurements and station temporal discontinuities by using spatiotemporal interpolation and neighboring observation collaborative interpolation to fill in missing data and adding missing value filling markers; spatiotemporal coordinate unification unifies the preset meteorological model grid, radar grid, and discrete station observations to the same geographic projection coordinate system and spatial grid resolution, completing full-domain grid matching; time dimension alignment unifies the time zone and time step of all data, completes time series pairing according to the rule of one-to-one correspondence between forecast time and measured time, eliminates unmatched isolated data, and achieves complete spatiotemporal synchronization between the initial weather forecast data and multi-source flood measured data.

[0025] After the above preprocessing, a standardized meteorological fusion dataset with spatiotemporal matching, no anomalies, and no missing data is formed, which is used for subsequent bi-objective training of the physical constraint flood forecasting large model.

[0026] Optionally, the meteorological and hydrological physical rules are embedded into a pre-defined large-scale flood forecasting model, including: The meteorological and hydrological physical rules are transformed into differentiable meteorological and hydrological residual operators. The encoder Transformer network is used as the preset flood forecasting large model. The network structure of the preset flood forecasting large model is as follows: input embedding layer, at least two multi-head self-attention coding modules, feedforward neural network, and forecast output head. A modular physical constraint layer is added between the multi-head self-attention encoding module and the feedforward neural network, and the differentiable meteorological and hydrological residual operator is built into the physical constraint layer.

[0027] In this embodiment, by extracting the hydrological and meteorological patterns of no precipitation and no flooding, and the matching of rainfall intensity and inundation level, the objective hydrological and meteorological patterns described in words are converted into differentiable mathematical residual operators, ensuring that the operators support gradient backpropagation and can participate in model training and optimization.

[0028] Furthermore, a standard Transformer network with only encoder structure is built as the preset base model. At this time, the network only has the ability to fit pure data and has no domain physical constraints.

[0029] Furthermore, an independent modular physical constraint layer is added between the output end of the multi-head self-attention module and the input end of the feedforward neural network in the preset flood forecasting model. All differentiable physical residual operators are embedded into the constraint layer. After the embedding is completed, the physical rules participate in each forward feature correction and loss calculation of the preset flood forecasting model, resulting in a basic flood forecasting model to be trained with physical priors.

[0030] It should be noted that in the forward propagation process of the pre-set flood forecasting model with embedded physical rules constructed in this embodiment, the spatiotemporal latent features output by the multi-head self-attention encoding module are input into the physical constraint layer. The built-in residual operator calculates the physical violation residual and corrects the latent features. After correction, the features are input into the feedforward neural network to complete the subsequent forecast inference. At the same time, the physical residuals output by the physical constraint layer can be used to construct the physical rule loss, so as to achieve the dual objective of joint training of data fitting accuracy and physical rule compliance rate.

[0031] Optionally, the preset flood forecasting model is trained based on the historical initial meteorological forecast data and historical multi-source flood measured data to obtain the target flood forecasting model, including: The initial rainfall intensity and initial inundation depth within the historical period are determined based on the historical initial meteorological forecast data, and the historical measured rainfall intensity and historical measured inundation depth within the historical period are determined based on the historical multi-source flood measured data. The initial rainfall intensity, initial inundation depth, historical measured rainfall intensity, and historical measured inundation depth at the same moment in the same grid and historical period are paired to construct a historical meteorological fusion time series sample. The historical meteorological fusion time series samples are divided into training set, test set and validation set. The preset flood forecasting model is trained based on the training set. The initial rainfall intensity and initial inundation depth are used as input features of the preset flood forecasting model, and the historical measured rainfall intensity and historical measured inundation depth are used as supervision labels of the preset flood forecasting model. The test set is input into the pre-trained flood forecasting model to obtain the predicted rainfall intensity and predicted inundation depth corresponding to the test set. The data fitting loss and physical rule loss are determined based on the test set. The data fitting loss is the error between the predicted rainfall intensity and predicted inundation depth of the test set and the historical measured rainfall intensity and historical measured inundation depth. The physical rule loss is the batch average physical residual output by the physical constraint layer. The network parameters of the preset flood forecasting model are updated based on the data fitting loss and physical rule loss. The process continues iterating until the data fitting loss and physical rule loss of the validation set are less than a preset threshold, thus obtaining the target flood forecasting large model.

[0032] In this embodiment, the initial rainfall intensity is the grid rainfall intensity initially predicted by the preset meteorological model, and the initial inundation depth is the grid inundation depth initially calculated by the preset meteorological model. Historical measured rainfall intensity and historical measured inundation depth are the actual observation values ​​collected by radar, meteorological stations, and hydrological stations. The data fitting loss is the error between the predicted rainfall intensity and predicted inundation depth output by the preset flood forecasting model based on the test set and the actual measured rainfall intensity and measured inundation depth, respectively, representing the degree to which the model's prediction results match the actual situation. The physical rule loss is the physical violation residual automatically calculated by the physical constraint layer within the preset flood forecasting model for each batch of prediction data, representing the degree to which the prediction results conform to natural hydrological and meteorological laws. The target flood forecasting model is the preset flood forecasting model after training.

[0033] In this embodiment, a time-series grid rainfall forecast field is extracted based on the historical initial weather forecast data output by the preset meteorological model, and the initial inundation depth of the grid is calculated based on the preset meteorological model. The rainfall intensity and initial inundation depth predicted by the preset meteorological model are used as input features of the preset flood forecasting model. Based on historical radar / meteorological station / hydrological station multi-source measured data, the measured rainfall intensity and measured inundation depth are extracted as supervision labels for the training process of the preset flood forecasting model.

[0034] Furthermore, based on the feature and label combinations completed by all spatiotemporal matching, historical meteorological fusion time-series samples are generated in batches. Further, all historical meteorological fusion time-series samples can be divided into training, testing, and validation sets according to a set ratio. The training set is used for the main process iterative training of the pre-set flood forecasting model, driving the model to learn the pre-set meteorological model bias, the spatiotemporal patterns of rainfall and water accumulation, and the meteorological, hydrological, and physical constraints. The testing set is used to uniformly calculate the data fitting loss and physical rule loss after each training round, serving as the basis for model parameter updates. The validation set is used to monitor the convergence status of the pre-set flood forecasting model after each training round and determine whether training has stopped.

[0035] Furthermore, the training set is batch-read and fused with time-series samples. The grid-based meteorological model forecast field (initial rainfall intensity, initial wind field, and initial inundation depth) is fed into the input embedding layer of the pre-set flood forecasting large model and mapped as high-dimensional spatiotemporal latent features. Then, the model's multi-layer multi-head self-attention encoding module extracts the correlation features of rainfall and water accumulation across the entire grid and between preceding and subsequent time series, and mines the long-term systematic deviation patterns of the pre-set meteorological model. Within the pre-set flood forecasting large model, the rainfall intensity and water depth variables in the features are separated, and the built-in differentiable residual operator is called to calculate the physical violation residual of a single grid. The abnormal spatiotemporal features are corrected according to the magnitude of the residuals, and the corrected features that conform to physical laws are output. The corrected features are fed into a feedforward neural network for nonlinear transformation, and finally, two types of predicted values ​​are output through the forecast output head: predicted rainfall intensity and predicted inundation depth.

[0036] Furthermore, the forecast fields (initial rainfall intensity, initial wind field, and initial inundation depth) of the pre-set meteorological model in the test set are input into the trained pre-set flood forecasting model to obtain the predicted rainfall intensity and predicted inundation depth corresponding to the test set. The mean square error is calculated by comparing these values ​​with the historical measured rainfall intensity and historical measured inundation depth in the test set samples. The larger the error value, the greater the deviation between the model prediction result and the actual observation, and the lower the data fitting accuracy. At the same time, the physical residuals of all grids in a single batch output by the physical constraint layer are read, and the average value is calculated to obtain the batch average physical residual, which is the physical rule loss. The larger the residual value, the more serious the violation of the hydrological and meteorological law of no flooding without precipitation and rainfall intensity matching water accumulation increase in the model output result.

[0037] Furthermore, a joint total loss can be constructed by weighting and summing the data fitting loss and the physical rule loss. The weights of the data fitting loss and the physical rule loss can be flexibly set to represent the degree of emphasis on data accuracy and physical compliance during model optimization.

[0038] Furthermore, with minimizing the total loss as the optimization objective, gradient backpropagation is performed to automatically solve the gradient of the total loss with respect to all learnable weights of the model; the parameters of the entire network are updated simultaneously, including all weight matrices of the input embedding layer, multi-head self-attention encoding module, physical constraint layer, feedforward neural network, and forecast output head. Then, parameter updates are performed to achieve dual optimization objectives: reducing data fitting loss, narrowing the deviation between predicted rainfall intensity and predicted inundation depth and actual observations, and improving forecast accuracy. During the model parameter update process, physical rule loss is also reduced to decrease abnormal predictions that violate objective hydrological laws, such as water accumulation in the absence of rain or heavy flooding in light rain, thereby improving forecast reliability.

[0039] In this embodiment, the entire process of forward prediction on the training set, calculation of double loss on the test set, and backward update of network parameters is repeated for multiple iterations. After each iteration, the validation set data fitting loss and validation set physical rule loss are calculated independently using validation set samples. After multiple iterations, if the values ​​of the validation set data fitting loss and physical rule loss remain basically stable and no longer decrease significantly, then training is stopped, and all network weights and network structure of the model are saved to obtain a target flood forecasting model that balances data fitting accuracy and physical rule compliance.

[0040] In this embodiment, the training samples of the pre-set flood forecasting model are integrated with the forecast data of the pre-set meteorological model and the measured flood data from multiple sources. It specifically learns the bias of the pre-set meteorological model itself, and the correction effect is better than that of the general model trained with only observation data. The physical rules are transformed into differentiable operators and embedded in the network. The model participates in gradient updates throughout the training process, rather than simply filtering out outliers after training. In this embodiment, the pre-set flood forecasting model is jointly optimized by adopting dual-objective loss. It does not simply fit the observed true values, but forces the output to conform to the meteorological and hydrological physical laws. The convergence is monitored synchronously by the dual loss of the validation set, which ensures the fit between the flood prediction data of the final target flood forecasting model and the measured flood data, while improving the physical rationality of the predicted flood data.

[0041] S130. Input the initial meteorological forecast data of the target into the target flood forecast model and output the target flood forecast data of the target area for the target period.

[0042] In this embodiment, the target initial meteorological forecast data includes the gridded rainfall forecast field, wind field forecast field, initial inundation depth, and supporting auxiliary meteorological and temporal spatial information output by the preset meteorological model. The standardized target initial meteorological forecast data is input into the trained target flood forecast model. After input embedding, multi-head self-attention feature extraction, real-time verification and correction by the physical constraint layer, and feedforward network inference, the target flood forecast data adapted to the target area and target time period is output. The target flood forecast data includes the gridded predicted rainfall intensity and gridded predicted inundation depth after bias correction and physical rule constraints.

[0043] Furthermore, in practical applications, lightweight optimizations can be performed on the target flood forecast model to improve computational efficiency and determine minute-level urban flood forecast results based on the target flood forecast data.

[0044] Optionally, after inputting the initial meteorological forecast data of the target into the target flood forecasting model and outputting the flood forecast data of the target area, the following steps are taken: Determine the measured multi-source flood data for the target area and target time; Based on the target flood forecast data and the target multi-source flood measured data, the flood forecast error is determined; Based on the flood prediction error, the parameters of the target flood forecasting model are corrected.

[0045] In this embodiment, the target multi-source flood measured data are the true observation values ​​collected synchronously in the target area and the target time period, including the radar measured rainfall intensity and the hydrological station measured inundation depth during the target time period.

[0046] The flood prediction error is the data error between the target flood forecast data and the target multi-source measured flood data. It is used to quantify the deviation of the current flood forecast data. The flood prediction error can be the difference between the predicted inundation depth and the measured inundation depth within the same grid and the same time period, or the difference between the predicted rainfall intensity and the measured rainfall intensity within the same grid and the same time period, or the total error of the predicted rainfall intensity, predicted inundation depth, and measured rainfall intensity and measured inundation depth within the same grid and the same time period. Specifically, the global grid flood prediction error can be obtained through error calculation formulas (such as mean square error and single-point absolute error). This error directly reflects the magnitude and distribution pattern of the flood forecast deviation of the current target flood forecast model. Simultaneously, the rainfall intensity error can be calculated as an auxiliary reference. Furthermore, based on this flood prediction error, incremental training samples are constructed, and the target flood forecast model is incrementally fine-tuned using a dual-objective loss method. This completes the correction of the network parameters of the target flood forecast model, achieving continuous iterative improvement of the model's forecast accuracy. Furthermore, the error patterns of the preset meteorological model extracted from the target flood forecasting model can be fed back to the preset meteorological model to optimize the basic forecast parameters of the preset meteorological model, such as the initial rainfall intensity and the initial inundation depth. The optimized preset meteorological model and the target flood forecasting model can then re-infer each other in a collaborative manner, thereby achieving a continuous improvement in the accuracy of the fusion forecast.

[0047] The technical solution of this invention embeds meteorological and hydrological physical rules into a preset flood forecasting model, integrates historical multi-source flood measurement data corresponding to historical time periods with historical initial meteorological forecast data output by the preset meteorological model, constructs training samples to train the preset flood forecasting model, obtains a target flood forecasting model, and uses the target initial meteorological forecast data for the target time period as a basis to perform data correction and meteorological and hydrological physical rule verification on the target initial meteorological forecast data, thereby obtaining the target flood forecast data for the target time period, improving the accuracy and physical rationality of the flood forecast data.

[0048] Example 2 Figure 2 This is a schematic diagram of a flood forecast data determination device according to Embodiment 2 of the present invention. This embodiment is applicable to flood forecast data determination. The flood forecast data determination device can be implemented in hardware and / or software, and can be configured in any electronic device with network communication and computing capabilities. Figure 2 As shown, the device includes: The data determination module 310 is used to determine the historical multi-source flood measured data and the historical initial meteorological forecast data output by the preset meteorological model for at least two historical time periods in the target area, the target initial meteorological forecast data output by the preset meteorological model for the target time period, and the meteorological and hydrological physical rules adapted to the urban flood forecast scenario; the meteorological and hydrological physical rules include at least the no-flood constraint rule and the rainfall intensity and inundation water level matching constraint rule. The target model determination module 320 is used to embed the meteorological and hydrological physical rules into the preset flood forecasting model, and to train the preset flood forecasting model based on the historical initial meteorological forecast data and historical multi-source flood measured data to obtain the target flood forecasting model. The flood forecast data determination module 330 is used to input the initial meteorological forecast data of the target into the target flood forecast big model and output the target flood forecast data of the target area for the target period.

[0049] Optionally, before embedding the meteorological and hydrological physical rules into a preset flood forecasting model and training the preset flood forecasting model based on the historical initial meteorological forecast data and historical multi-source flood measurement data to obtain the target flood forecasting model, the following steps are included: The historical multi-source flood measured data and historical initial meteorological forecast data are preprocessed, including cleaning and denoising, missing value filling, spatiotemporal coordinate unification, and time dimension alignment.

[0050] Optionally, the meteorological and hydrological physical rules are embedded into a pre-defined large-scale flood forecasting model, including: The meteorological and hydrological physical rules are transformed into differentiable meteorological and hydrological residual operators. The encoder Transformer network is used as the preset flood forecasting large model. The network structure of the preset flood forecasting large model is as follows: input embedding layer, at least two multi-head self-attention coding modules, feedforward neural network, and forecast output head. A modular physical constraint layer is added between the multi-head self-attention encoding module and the feedforward neural network, and the differentiable meteorological and hydrological residual operator is built into the physical constraint layer.

[0051] Optionally, the preset flood forecasting model is trained based on the historical initial meteorological forecast data and historical multi-source flood measured data to obtain the target flood forecasting model, including: The initial rainfall intensity and initial inundation depth within the historical period are determined based on the historical initial meteorological forecast data, and the historical measured rainfall intensity and historical measured inundation depth within the historical period are determined based on the historical multi-source flood measured data. The initial rainfall intensity, initial inundation depth, historical measured rainfall intensity, and historical measured inundation depth at the same moment in the same grid and historical period are paired to construct a historical meteorological fusion time series sample. The historical meteorological fusion time series samples are divided into training set, test set and validation set. The preset flood forecasting model is trained based on the training set. The initial rainfall intensity and initial inundation depth are used as input features of the preset flood forecasting model, and the historical measured rainfall intensity and historical measured inundation depth are used as supervision labels of the preset flood forecasting model. The test set is input into the pre-trained flood forecasting model to obtain the predicted rainfall intensity and predicted inundation depth corresponding to the test set. The data fitting loss and physical rule loss are determined based on the test set. The data fitting loss is the error between the predicted rainfall intensity and predicted inundation depth of the test set and the historical measured rainfall intensity and historical measured inundation depth. The physical rule loss is the batch average physical residual output by the physical constraint layer. The network parameters of the preset flood forecasting model are updated based on the data fitting loss and physical rule loss. The process continues iterating until the data fitting loss and physical rule loss of the validation set are less than a preset threshold, thus obtaining the target flood forecasting large model.

[0052] Optionally, after inputting the initial meteorological forecast data of the target into the target flood forecasting model and outputting the target flood forecast data of the target area, the process includes: Determine the measured multi-source flood data for the target area and time period; Based on the target flood forecast data and the target multi-source flood measured data, the flood forecast error is determined; Based on the flood prediction error, the parameters of the target flood forecasting model are corrected.

[0053] Optionally, the multi-source flood measurement data includes at least radar measurement data, meteorological station measurement data, and hydrological station measurement water level data. The radar measurement data includes radar-measured instantaneous rainfall rate, measured rainfall in each time period, and average rainfall over the basin area. The meteorological station measurement data includes meteorological station measured air temperature, meteorological station measured relative humidity, and meteorological station measured surface air pressure. The hydrological station measurement water level data includes measured water level depth, measured water level rise, measured water level change rate, and measured instantaneous flow rate at the river cross-section.

[0054] Optionally, the initial weather forecast data includes initial rainfall forecast data, initial wind field forecast data, and initial inundation depth data output by a preset weather model. The rainfall forecast data includes instantaneous rainfall rate, cumulative rainfall for each time period, rainfall type classification, and rainfall duration. The wind field forecast data includes horizontal wind speed at a preset height above the ground, horizontal wind direction azimuth, and instantaneous wind speed.

[0055] The technical solution of this invention embeds meteorological and hydrological physical rules into a preset flood forecasting model, integrates historical multi-source flood measurement data corresponding to historical time periods with historical initial meteorological forecast data output by the preset meteorological model, constructs training samples to train the preset flood forecasting model, obtains a target flood forecasting model, and uses the target initial meteorological forecast data for the target time period as a basis to perform data correction and meteorological and hydrological physical rule verification on the target initial meteorological forecast data, thereby obtaining the target flood forecast data for the target time period, improving the accuracy and physical rationality of the flood forecast data.

[0056] The flood forecast data determination device provided in this embodiment of the invention can execute the flood forecast data determination method provided in any embodiment of the invention, and has the corresponding functional modules and beneficial effects of the method.

[0057] Example 3 Figure 3 A schematic diagram of an electronic device 10, which can be used to implement embodiments of the present invention, is shown. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as personal digital processors, cellular phones, smartphones, wearable devices (e.g., helmets, glasses, watches, etc.), and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely illustrative and are not intended to limit the implementation of the invention described and / or claimed herein.

[0058] like Figure 3As shown, the electronic device 10 includes at least one processor 11 and a memory, such as a read-only memory (ROM) 12 or a random access memory (RAM) 13, communicatively connected to the at least one processor 11. The memory stores computer programs executable by the at least one processor. The processor 11 can perform various appropriate actions and processes based on the computer program stored in the ROM 12 or loaded from storage unit 18 into the RAM 13. The RAM 13 can also store various programs and data required for the operation of the electronic device 10. The processor 11, ROM 12, and RAM 13 are interconnected via a bus 14. An input / output (I / O) interface 15 is also connected to the bus 14.

[0059] Multiple components in electronic device 10 are connected to I / O interface 15, including: input unit 16, such as keyboard, mouse, etc.; output unit 17, such as various types of displays, speakers, etc.; storage unit 18, such as disk, optical disk, etc.; and communication unit 19, such as network card, modem, wireless transceiver, etc. Communication unit 19 allows electronic device 10 to exchange information / data with other devices through computer networks such as the Internet and / or various telecommunications networks.

[0060] Processor 11 can be a variety of general-purpose and / or special-purpose processing components with processing and computing capabilities. Some examples of processor 11 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various special-purpose artificial intelligence (AI) computing chips, various processors running machine learning model algorithms, digital signal processors (DSPs), and any suitable processor, controller, microcontroller, etc. Processor 11 performs the various methods and processes described above, such as flood forecast data determination methods.

[0061] In some embodiments, the flood forecast data determination method may be implemented as a computer program tangibly contained in a computer-readable storage medium, such as storage unit 18. In some embodiments, part or all of the computer program may be loaded and / or installed on electronic device 10 via ROM 12 and / or communication unit 19. When the computer program is loaded into RAM 13 and executed by processor 11, one or more steps of the flood forecast data determination method described above may be performed. Alternatively, in other embodiments, processor 11 may be configured to perform the flood forecast data determination method by any other suitable means (e.g., by means of firmware).

[0062] In particular, according to embodiments of the present invention, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, embodiments of the present invention include a computer program product comprising a computer program carried on a non-transitory computer-readable medium, the computer program containing program code for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via communication unit 19, or installed from storage unit 18, or installed from ROM 12. When the computer program is executed by processor 11, it performs the functions defined in the methods of the embodiments of the present invention.

[0063] Various embodiments of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field-programmable gate arrays (FPGAs), application-specific integrated circuits (ASICs), application-specific standard products (ASSPs), system-on-a-chip (SoCs), complex programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various embodiments may include implementations in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which may be a dedicated or general-purpose programmable processor, capable of receiving data and instructions from a storage system, at least one input device, and at least one output device, and transmitting data and instructions to the storage system, the at least one input device, and the at least one output device.

[0064] Computer programs used to implement the methods of the present invention may be written in any combination of one or more programming languages. These computer programs may be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing device, such that when executed by the processor, the computer programs cause the functions / operations specified in the flowcharts and / or block diagrams to be performed. The computer programs may be executed entirely on a machine, partially on a machine, or as a standalone software package, partially on a machine and partially on a remote machine, or entirely on a remote machine or server.

[0065] In the context of this invention, a computer-readable storage medium can be a tangible medium that may contain or store a computer program for use by or in conjunction with an instruction execution system, apparatus, or device. A computer-readable storage medium may include, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination thereof. Alternatively, a computer-readable storage medium may be a machine-readable signal medium. More specific examples of machine-readable storage media include electrical connections based on one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fibers, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof.

[0066] To provide interaction with a user, the systems and techniques described herein can be implemented on an electronic device having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user; and a keyboard and pointing device (e.g., a mouse or trackball) through which the user provides input to the electronic device. Other types of devices can also be used to provide interaction with the user; for example, feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including sound input, voice input, or tactile input).

[0067] The systems and technologies described herein can be implemented in computing systems that include backend components (e.g., as data servers), or middleware components (e.g., application servers), or frontend components (e.g., user computers with graphical user interfaces or web browsers through which users can interact with implementations of the systems and technologies described herein), or any combination of such backend, middleware, or frontend components. The components of the system can be interconnected via digital data communication of any form or medium (e.g., communication networks). Examples of communication networks include local area networks (LANs), wide area networks (WANs), blockchain networks, and the Internet.

[0068] A computing system can include clients and servers. Clients and servers are generally located far apart and typically interact through communication networks. The client-server relationship is created by computer programs running on the respective computers and having a client-server relationship with each other. The server can be a cloud server, also known as a cloud computing server or cloud host, which is a hosting product within the cloud computing service system to address the shortcomings of traditional physical hosts and VPS services, such as high management difficulty and weak business scalability.

[0069] It should be understood that the various forms of processes shown above can be used, with steps reordered, added, or deleted. For example, the steps described in this invention can be executed in parallel, sequentially, or in different orders, as long as the desired result of the technical solution of this invention can be achieved, and this is not limited herein.

[0070] The specific embodiments described above do not constitute a limitation on the scope of protection of this invention. Those skilled in the art should understand that various modifications, combinations, sub-combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this invention should be included within the scope of protection of this invention.

Claims

1. A method for determining flood forecast data, characterized in that, include: The system determines at least two historical periods corresponding to the target area, including historical multi-source flood measured data and historical initial meteorological forecast data output by a preset meteorological model, target initial meteorological forecast data output by a preset meteorological model for the target period, and meteorological, hydrological, and physical rules adapted to urban flood forecasting scenarios. The meteorological, hydrological, and physical rules include at least a no-flood constraint rule and a rainfall intensity-inundation level matching constraint rule. The meteorological and hydrological physical rules are embedded into a preset flood forecasting model, and the preset flood forecasting model is trained based on the historical initial meteorological forecast data and historical multi-source flood measured data to obtain the target flood forecasting model. The initial meteorological forecast data of the target area is input into the target flood forecast model, and the target flood forecast data of the target area for the target period is output.

2. The method according to claim 1, characterized in that, Before embedding the meteorological and hydrological physical rules into a preset flood forecasting model, and training the preset flood forecasting model based on the historical initial meteorological forecast data and historical multi-source flood measurement data to obtain the target flood forecasting model, the following steps are included: The historical multi-source flood measured data and historical initial meteorological forecast data are preprocessed, including cleaning and denoising, missing value filling, spatiotemporal coordinate unification, and time dimension alignment.

3. The method according to claim 1, characterized in that, Embedding the meteorological and hydrological physical rules into a pre-set large-scale flood forecasting model includes: The meteorological and hydrological physical rules are transformed into differentiable meteorological and hydrological residual operators. The encoder Transformer network is used as the preset flood forecasting large model. The network structure of the preset flood forecasting large model is as follows: input embedding layer, at least two multi-head self-attention coding modules, feedforward neural network, and forecast output head. A modular physical constraint layer is added between the multi-head self-attention encoding module and the feedforward neural network, and the differentiable meteorological and hydrological residual operator is built into the physical constraint layer.

4. The method according to claim 1, characterized in that, The pre-set flood forecasting model is trained based on the historical initial meteorological forecast data and historical multi-source flood measured data to obtain the target flood forecasting model, including: The initial rainfall intensity and initial inundation depth within the historical period are determined based on the historical initial meteorological forecast data, and the historical measured rainfall intensity and historical measured inundation depth within the historical period are determined based on the historical multi-source flood measured data. The initial rainfall intensity, initial inundation depth, historical measured rainfall intensity, and historical measured inundation depth at the same moment in the same grid and historical period are paired to construct a historical meteorological fusion time series sample. The historical meteorological fusion time series samples are divided into training set, test set and validation set. The preset flood forecasting model is trained based on the training set. The initial rainfall intensity and initial inundation depth are used as input features of the preset flood forecasting model, and the historical measured rainfall intensity and historical measured inundation depth are used as supervision labels of the preset flood forecasting model. The test set is input into the pre-trained flood forecasting model to obtain the predicted rainfall intensity and predicted inundation depth corresponding to the test set. The data fitting loss and physical rule loss are determined based on the test set. The data fitting loss is the error between the predicted rainfall intensity and predicted inundation depth of the test set and the historical measured rainfall intensity and historical measured inundation depth. The physical rule loss is the batch average physical residual output by the physical constraint layer. The network parameters of the preset flood forecasting model are updated based on the data fitting loss and physical rule loss. The process continues iterating until the data fitting loss and physical rule loss of the validation set are less than a preset threshold, thus obtaining the target flood forecasting large model.

5. The method according to claim 1, characterized in that, After inputting the initial meteorological forecast data of the target area into the target flood forecasting model and outputting the target flood forecasting data for the target area, the following steps are taken: Determine the measured multi-source flood data for the target area and time period; Based on the target flood forecast data and the target multi-source flood measured data, the flood forecast error is determined; Based on the flood prediction error, the parameters of the target flood forecasting model are corrected.

6. The method according to claim 1, characterized in that, The multi-source flood measurement data includes at least radar measurement data, meteorological station measurement data, and hydrological station measurement water level data. The radar measurement data includes radar-measured instantaneous rainfall rate, measured rainfall amount for each time period, and average rainfall amount over the basin area. The meteorological station measurement data includes meteorological station measured air temperature, meteorological station measured relative humidity, and meteorological station measured surface air pressure. The hydrological station measurement water level data includes measured water level depth, measured water level rise, measured water level change rate, and measured instantaneous flow rate at the river cross-section.

7. The method according to claim 1, characterized in that, The initial meteorological forecast data includes initial rainfall forecast data, initial wind field forecast data, and initial inundation depth data output by a preset meteorological model. The rainfall forecast data includes instantaneous rainfall rate, cumulative rainfall for each time period, rainfall type classification, and rainfall duration. The wind field forecast data includes horizontal wind speed at a preset height above the ground, horizontal wind direction azimuth, and instantaneous wind speed.

8. A device for determining flood forecast data, characterized in that, include: The data determination module is used to determine the historical multi-source flood measured data and the historical initial meteorological forecast data output by the preset meteorological model for at least two historical time periods in the target area, the target initial meteorological forecast data output by the preset meteorological model for the target time period, and the meteorological and hydrological physical rules adapted to the urban flood forecast scenario; the meteorological and hydrological physical rules include at least the no-flood constraint rule and the rainfall intensity and inundation water level matching constraint rule; The target model determination module is used to embed the meteorological and hydrological physical rules into the preset flood forecasting model, and to train the preset flood forecasting model based on the historical initial meteorological forecast data and historical multi-source flood measured data to obtain the target flood forecasting model. The flood forecast data determination module is used to input the initial meteorological forecast data of the target into the target flood forecast big model and output the target flood forecast data of the target area for the target period.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the flood forecast data determination method as described in any one of claims 1-7.

10. A storage medium for storing computer-executable instructions, characterized in that, The computer-executable instructions, when executed by a computer processor, are used to perform the flood forecast data determination method as described in any one of claims 1-7.

Images