Reservoir level determination method and device, program product and electronic equipment

Abstract

The invention discloses a reservoir level determination method and device, a program product and electronic equipment, and relates to the technical field of water level monitoring, the method comprises the following steps: collecting multi-source remote sensing data corresponding to a target reservoir, the multi-source remote sensing data being used for representing surface information of a drainage basin to which the target reservoir belongs; preprocessing the multi-source remote sensing data to obtain target remote sensing data corresponding to the target reservoir; performing inversion based on the target remote sensing data to obtain soil water content data of the target reservoir; the water level of the target reservoir is determined based on the soil water content data of the target reservoir and the target precipitation data, and the target precipitation data is used for representing precipitation information of a drainage basin to which the target reservoir belongs. The technical problem that the accuracy of water level detection of a runoff-free reservoir based on the prior art is low is solved.

Description

Technical Field

[0001] This application relates to the field of water level monitoring technology, and more specifically, to a method, apparatus, program product, and electronic device for determining the water level of a reservoir. Background Technology

[0002] Global reservoir systems occupy a central position in water resource management. As of 2024, there were 61,988 dams with a capacity of over 3 million cubic meters worldwide. These water conservancy facilities play a key role in flood control, agricultural irrigation, urban water supply, and clean energy production. For example, in 2023, the global hydropower capacity reached 1,412 gigawatts, accounting for about 15% of global power generation. However, reservoir management faces the following challenges: climate change and human activities have led to highly dynamic changes in water area, water level, and water storage. More than 95% of small and medium-sized river basins worldwide lack basic monitoring data, forming a long-standing scientific problem of runoff prediction in areas without data.

[0003] Furthermore, the severity of flood disasters further highlights the urgency of breakthroughs in reservoir water level technology. Since the beginning of this century, the frequency of catastrophic floods worldwide has surged by 134%. In complex geographical environments, such as water basins affected by factors like inter-regional rainfall, there is a risk of water wastage based on traditional manual experience-based scheduling methods.

[0004] In existing technologies, reservoir water level monitoring mainly relies on ground monitoring stations to estimate water level changes by directly or indirectly measuring runoff data. However, this reliance becomes more vulnerable to reservoirs without runoff monitoring data. Due to the lack of monitoring stations, traditional methods of collecting runoff data become ineffective, making it impossible to effectively verify existing hydrological models and significantly reducing the accuracy of reservoir water level prediction. This results in the technical problem of low accuracy in detecting water levels in reservoirs without runoff monitoring based on existing technologies.

[0005] There is currently no effective solution to the above problems. Summary of the Invention

[0006] This application provides a method, apparatus, program product, and electronic device for determining reservoir water level, to at least solve the technical problem of low accuracy in water level detection of reservoirs without runoff based on existing technologies.

[0007] According to one aspect of this application, a method for determining the water level of a reservoir is provided, comprising: collecting multi-source remote sensing data corresponding to a target reservoir, wherein the multi-source remote sensing data is used to characterize the surface information of the watershed to which the target reservoir belongs; preprocessing the multi-source remote sensing data to obtain target remote sensing data corresponding to the target reservoir, wherein the preprocessing is at least used to unify the resolution scale corresponding to the multi-source remote sensing data; inverting the soil moisture content data of the target reservoir based on the target remote sensing data; and determining the water level of the target reservoir based on the soil moisture content data and target precipitation data, wherein the target precipitation data is used to characterize the precipitation information of the watershed to which the target reservoir belongs.

[0008] Optionally, multi-source remote sensing data corresponding to the target reservoir is collected, including: dividing the watershed boundary of the target reservoir based on the land data corresponding to the target reservoir to obtain the watershed to which the target reservoir belongs, wherein the land data includes at least the land type and land use of the watershed to which the target reservoir belongs; collecting microwave data and optical data corresponding to the watershed to which the target reservoir belongs, wherein the microwave data is surface data obtained by detecting microwave signals and the optical data is surface data obtained by detecting light wave signals; and using the microwave data and optical data as multi-source remote sensing data corresponding to the target reservoir.

[0009] Optionally, preprocessing the multi-source remote sensing data to obtain target remote sensing data corresponding to the target reservoir includes: spatiotemporally aligning the multi-source remote sensing data based on a preset resolution to obtain first remote sensing data with consistent resolution scale, wherein the preset resolution includes at least a preset temporal resolution and a preset spatial resolution; performing data cleaning and data correction on the first remote sensing data to obtain second remote sensing data corresponding to the target reservoir, wherein data cleaning is used to remove optical data in the first remote sensing data with a cloud obscuration rate greater than a preset obscuration rate, and data correction is used to perform radiometric correction on microwave data in the first remote sensing data; detecting the data quality score corresponding to the second remote sensing data, wherein the data quality score is used to characterize at least the radiometric accuracy and image geometric accuracy of the second remote sensing data; generating a reacquisition signal when the data quality score is less than a preset score, wherein the reacquisition signal is used to reacquire the multi-source remote sensing data corresponding to the target reservoir; and using the second remote sensing data as the target remote sensing data when the data quality score is greater than or equal to the preset score.

[0010] Optionally, the soil moisture content data of the target reservoir is obtained by inverting the target remote sensing data, including: inputting the target remote sensing data into the target inversion model, wherein the target inversion model includes at least an inversion feature extraction layer and an inversion physical constraint layer, and the target inversion model is a neural network model trained based on historical soil moisture content data and historical remote sensing data corresponding to P reservoirs, where P is a positive integer; extracting features from the target remote sensing data through the inversion feature extraction layer to obtain L surface features, where L is a positive integer, and the L surface features are used to characterize at least the electromagnetic wave energy intensity, vegetation growth status, and surface temperature information of the watershed to which the target reservoir belongs; determining the soil moisture content sequence of the target reservoir based on the L surface features; detecting whether the soil moisture content sequence meets the preset soil moisture constraint conditions through the inversion physical constraint layer; and using the soil moisture content sequence as the soil moisture content data of the target reservoir if the soil moisture content sequence meets the preset soil moisture constraint conditions.

[0011] Optionally, determining the water level of the target reservoir based on soil moisture content data and target precipitation data includes: inputting the soil moisture content data and target precipitation data into a target detection model, wherein the target detection model includes at least a long short-term memory network layer, an attention mechanism module, and a water level physical constraint layer; extracting features from the soil moisture content data and target precipitation data through the long short-term memory network layer to obtain the soil moisture content features and precipitation features corresponding to the target reservoir; determining the feature weights corresponding to the soil moisture content features and precipitation features based on the climate conditions of the watershed to which the target reservoir belongs through the attention mechanism module; determining the reservoir water volume change rate of the target reservoir based on the soil moisture content features, precipitation features, and the feature weights corresponding to the soil moisture content features and precipitation features; detecting whether the reservoir water volume change rate meets the preset reservoir water volume constraint conditions through the water level physical constraint layer; and determining the water level of the target reservoir based on the reservoir water volume change rate if the reservoir water volume change rate meets the preset reservoir water volume constraint conditions.

[0012] Optionally, the training steps of the target detection model include: collecting historical water level data, historical soil moisture content data, and historical precipitation data from Q reservoirs without runoff, where Q is a positive integer; dividing the historical water level data, historical soil moisture content data, and historical precipitation data from the Q reservoirs into a training set and a validation set; iteratively training the initial model based on samples in the training set to obtain a training model; determining the loss function value corresponding to the training model based on samples in the validation set, wherein the loss function value is determined based on the water level prediction error, physical constraint loss, and regularization term in the attention mechanism module of the training model; and using the training model obtained from the last training as the target detection model if the loss function value is less than or equal to a preset function value.

[0013] Optionally, after determining the water level of the target reservoir based on the soil moisture content data and the target precipitation data, the method for determining the reservoir water level further includes: generating early warning information when the water level of the target reservoir is higher than or equal to a preset water level, wherein the early warning information is used to issue flood warnings for the watershed to which the target reservoir belongs.

[0014] According to another aspect of this application, a device for determining the water level of a reservoir is also provided, comprising: a data acquisition unit for acquiring multi-source remote sensing data corresponding to a target reservoir, wherein the multi-source remote sensing data is used to characterize the surface information of the watershed to which the target reservoir belongs; a data preprocessing unit for preprocessing the multi-source remote sensing data to obtain target remote sensing data corresponding to the target reservoir, wherein the preprocessing is at least used to unify the resolution scale corresponding to the multi-source remote sensing data; a soil moisture content data determination unit for inverting the soil moisture content data of the target reservoir based on the target remote sensing data; and a water level determination unit for determining the water level of the target reservoir based on the soil moisture content data and target precipitation data of the target reservoir, wherein the target precipitation data is used to characterize the precipitation information of the watershed to which the target reservoir belongs.

[0015] According to another aspect of this application, a computer program product is also provided, which stores a computer program, wherein, when the computer program is running, it controls the computer program product to execute any of the above-mentioned methods for determining the reservoir water level.

[0016] According to another aspect of this application, an electronic device is also provided, wherein the electronic device includes one or more processors and a memory for storing one or more programs, wherein when the one or more programs are executed by the one or more processors, the one or more processors implement the method for determining the reservoir water level as described above.

[0017] In this application, firstly, multi-source remote sensing data corresponding to the target reservoir is collected. The multi-source remote sensing data is used to characterize the surface information of the watershed to which the target reservoir belongs. Then, the multi-source remote sensing data is preprocessed to obtain the target remote sensing data corresponding to the target reservoir. The preprocessing is used at least to unify the resolution scale of the multi-source remote sensing data. Then, the soil moisture content data of the target reservoir is obtained by inversion based on the target remote sensing data. Subsequently, the water level of the target reservoir is determined based on the soil moisture content data and the target precipitation data. The target precipitation data is used to characterize the precipitation information of the watershed to which the target reservoir belongs.

[0018] As can be seen from the above, this application determines the soil moisture content data corresponding to the watershed of the reservoir with no runoff based on multi-source remote sensing data, and uses the soil moisture content data to replace the runoff monitoring data in traditional technology to predict the reservoir water level, thereby overcoming the technical bottleneck of scarce runoff monitoring data in the existing technology, and thus achieving the purpose of accurately predicting the reservoir water level.

[0019] Specifically, this application first unifies the spatiotemporal resolution of multi-source remote sensing data by preprocessing the pre-collected multi-source remote sensing data, thereby improving the accuracy of the acquired reservoir soil moisture content data and effectively filling the gap in reservoir water level monitoring in areas without monitoring stations. Then, this application determines the reservoir water level based on a two-layer factor of reservoir soil moisture content data and reservoir precipitation data, thereby achieving the technical effect of improving the timeliness and accuracy of reservoir water level detection results, and thus solving the technical problem of low accuracy in water level detection of reservoirs without runoff based on existing technologies. Attached Figure Description

[0020] The accompanying drawings, which are included to provide a further understanding of this application and form part of this application, illustrate exemplary embodiments and are used to explain this application, but do not constitute an undue limitation of this application. In the drawings:

[0021] Figure 1 This is a flowchart of an optional method for determining the water level of a reservoir according to an embodiment of this application;

[0022] Figure 2 This is an architecture diagram of an optional target detection model according to an embodiment of this application;

[0023] Figure 3 A flowchart of an optional reservoir water level forecasting method based on multi-source remote sensing data and a physically constrained neural network according to an embodiment of this application;

[0024] Figure 4 A schematic diagram of an optional reservoir water level determination device according to an embodiment of this application;

[0025] Figure 5 This is a structural block diagram of an electronic device according to an embodiment of this application. Detailed Implementation

[0026] To enable those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present application, and not all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative effort should fall within the scope of protection of the present application.

[0027] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this application 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 this application 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 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.

[0028] It should also be noted that all information and data (including but not limited to information used for display and analysis) involved in this application are authorized by the user or fully authorized by all parties. For example, if there is an interface between this system and the relevant user or organization, before obtaining the relevant information, it is necessary to send a request to the aforementioned user or organization through the interface, and obtain the relevant information only after receiving consent from the aforementioned user or organization.

[0029] Furthermore, the collection, storage, use, processing, transmission, provision, disclosure, and application of relevant information and data involved in this application all comply with the relevant laws, regulations, and standards of the relevant regions, and necessary confidentiality measures have been taken. This application does not violate public order and good morals. In addition, this application provides a corresponding operation entry point for users to choose to agree to or refuse authorization. If the user chooses to refuse authorization, the corresponding expert decision-making process will be initiated.

[0030] First, some nouns or terms that appear in the description of the embodiments of this application shall be interpreted as follows:

[0031] Reservoirs with no runoff: These are reservoirs that have no or only sparse runoff (i.e., water flowing into or out of the reservoir) monitoring stations or monitoring data.

[0032] To address the challenges in monitoring non-runoff reservoirs in existing technologies, this application provides a technical solution that combines remote sensing as a substitute for monitoring, physical constraint neural network fusion, and dynamic weight allocation.

[0033] (1) Remote sensing as a substitute for monitoring: Soil moisture content is retrieved using multi-source remote sensing (microwave data + optical data) as an indirect proxy variable for runoff (soil moisture content directly affects the runoff generation and confluence process).

[0034] (2) Physically constrained neural network: Hydrological physical equations are embedded in the LSTM (Long Short-Term Memory) network to force the output of the neural network to be constrained within the physical feasible domain, thus avoiding the failure of the pure data-driven model when data is scarce.

[0035] (3) Attention mechanism innovation: Design a dual-stream attention module to dynamically calculate the temporal weights of precipitation and soil moisture content, and solve the regional adaptability problem of "precipitation dominance in arid areas / soil moisture content dominance in humid areas".

[0036] The present invention will now be described in detail with reference to various embodiments.

[0037] Example 1

[0038] According to an embodiment of this application, an embodiment of a method for determining the water level of a reservoir is provided. It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions. Furthermore, although a logical order is shown in the flowchart, in some cases, the steps shown or described may be executed in a different order than that shown here.

[0039] This application provides a reservoir water level determination system (hereinafter referred to as the water level determination system) for executing the reservoir water level determination method in this application. Figure 1 This is a flowchart of an optional method for determining reservoir water level according to an embodiment of this application, such as... Figure 1 As shown, the method includes the following steps:

[0040] Step S101: Collect multi-source remote sensing data corresponding to the target reservoir, wherein the multi-source remote sensing data is used to characterize the surface information of the watershed to which the target reservoir belongs.

[0041] Optionally, multi-source remote sensing data refers to multiple remote sensing observation data corresponding to the watershed to which the target reservoir belongs, collected by satellites, aircraft or drones. The multi-source remote sensing data corresponding to the target reservoir is determined based on microwave datasets (e.g., SAR datasets (a dataset collected by a satellite equipped with a C-band synthetic aperture radar)) and optical datasets (e.g., MODIS datasets (a dataset collected by a satellite equipped with a high-spectral-resolution imaging spectrometer)) pre-collected by the water level determination system.

[0042] Optionally, the multi-source remote sensing data collected by the water level determination system can provide surface information of the target reservoir's watershed from different perspectives, making up for the limitations of a single data source. This improves the comprehensiveness and reliability of the reservoir surface information monitoring results. The remote sensing data has a wide coverage, fast update frequency, and high resolution, enabling long-term trend analysis of surface information of the target reservoir and its surrounding watershed. This fills the data gap for monitoring reservoirs in areas with scarce runoff data.

[0043] Step S102: Preprocess the multi-source remote sensing data to obtain the target remote sensing data corresponding to the target reservoir. The preprocessing is used at least to unify the resolution scale of the multi-source remote sensing data.

[0044] Optionally, preprocessing includes at least a series of operations such as spatiotemporal alignment, data cleaning, and data quality scoring and detection, thereby achieving the following objectives:

[0045] (1) Improve data quality: The preprocessing process eliminates noise interference factors in remote sensing data, improving the purity and usability of the data.

[0046] (2) Unified data scale: By resampling and interpolation, the resolution scale of different remote sensing data is unified, ensuring that the time series and spatial coverage of remote sensing data are targeted and real-time for the dynamic monitoring of the target reservoir, ensuring the processing accuracy of subsequent hydrological models (such as target inversion models), and improving the accuracy of soil moisture content predicted by hydrological models.

[0047] Step S103: Soil moisture content data of the target reservoir is obtained by inversion based on the target remote sensing data.

[0048] Optionally, the water level determination system uses remote sensing inversion technology to retrieve soil moisture data of the target reservoir based on the target remote sensing data. This soil moisture data reflects surface runoff and infiltration conditions, thus obtaining key variables for reservoir water level forecasting and achieving the following functions:

[0049] (1) Alternative runoff monitoring: Soil moisture content data can be used as an indirect indicator to reflect the water dynamics in the watershed when there is no surface runoff monitoring, and is especially suitable for hydrological monitoring of reservoirs without monitoring data.

[0050] (2) Enhanced model input: It provides a high-quality input variable related to reservoir water level prediction for the subsequent target detection model, thereby improving the water level detection accuracy of the target detection model.

[0051] Step S104: Determine the water level of the target reservoir based on the soil moisture content data and the target precipitation data, wherein the target precipitation data is used to characterize the precipitation information of the watershed to which the target reservoir belongs.

[0052] Optionally, the target precipitation data refers to the precipitation information of the watershed where the target reservoir is located, which can be directly obtained through a preset meteorological API interface.

[0053] Optionally, the water level determination system can utilize soil moisture content and precipitation data of the target reservoir, combined with a hydrological model (such as a target detection model) with an inversion physical constraint layer, to achieve accurate prediction of the actual water level of the target reservoir. The application of the inversion physical constraint layer in the target detection model ensures that the dynamic water level forecast conforms to hydrological laws, avoiding the distorted prediction results produced by the pure data-driven model in the prior art. In the application of the target detection model, the water level determination system comprehensively considers the influence of precipitation and soil moisture content of the watershed to which the target reservoir belongs on the reservoir water level. That is, the water level determination system enhances the adaptability of the target detection model to different reservoir climatic conditions and reservoir soil characteristics by dynamically allocating weights, thereby improving the accuracy of reservoir water level forecasting.

[0054] As can be seen from the above, this application determines the soil moisture content data corresponding to the watershed of the reservoir with no runoff based on multi-source remote sensing data, and uses the soil moisture content data to replace the runoff monitoring data in traditional technology to predict the reservoir water level, thereby overcoming the technical bottleneck of scarce runoff monitoring data in the existing technology, and thus achieving the purpose of accurately predicting the reservoir water level.

[0055] Specifically, this application first unifies the spatiotemporal resolution of multi-source remote sensing data by preprocessing the pre-collected multi-source remote sensing data, thereby improving the accuracy of the acquired reservoir soil moisture content data and effectively filling the gap in reservoir water level monitoring in areas without monitoring stations. Then, this application determines the reservoir water level based on a two-layer factor of reservoir soil moisture content data and reservoir precipitation data, thereby achieving the technical effect of improving the timeliness and accuracy of reservoir water level detection results, and thus solving the technical problem of low accuracy in water level detection of reservoirs without runoff based on existing technologies.

[0056] In one optional embodiment, in order to obtain multi-source remote sensing data corresponding to the target reservoir, the water level determination system first divides the watershed boundary of the target reservoir based on the land data corresponding to the target reservoir to obtain the watershed to which the target reservoir belongs. The land data includes at least the land type and land use of the watershed to which the target reservoir belongs. Then, the water level determination system collects microwave data and optical data corresponding to the watershed to which the target reservoir belongs. The microwave data is the surface data obtained by detecting microwave signals, and the optical data is the surface data obtained by detecting light wave signals. Then, the water level determination system uses the microwave data and optical data as the multi-source remote sensing data corresponding to the target reservoir.

[0057] Optionally, the water level determination system uses a DEM (Digital Elevation Model) to delineate the watershed boundary of the target reservoir based on land data. This accurately determines the reservoir's catchment area (e.g., the catchment area within a 10km radius of the dam site of the target reservoir is considered the watershed of the target reservoir). Based on the watershed boundary, the system can then crop the collected original multi-source remote sensing data to obtain the corresponding multi-source remote sensing data for the target reservoir. This improves the relevance of the input data for the subsequent target inversion model, avoids inversion prediction of soil moisture content data for the target reservoir based on irrelevant data, and thus improves the accuracy of the subsequently predicted soil moisture content data for the target reservoir.

[0058] Optionally, land data should include at least the land type and land use of the watershed to which the target reservoir belongs, such as forest, grassland, cultivated land, and urban areas. Consideration should be given to land type and land use, as different land types and land uses have different impacts on the hydrological evolution process. For example, forests can increase soil moisture, and land urbanization can increase runoff velocity. By comprehensively analyzing land data, the system can improve the accuracy and precision of the watershed boundary corresponding to the target reservoir determined by the water level determination system. This will help subsequent hydrological models to more accurately detect the hydrological characteristics of the watershed to which the target reservoir belongs and improve forecast accuracy.

[0059] Optionally, microwave data refers to a dataset of surface images obtained by using microwave signals to detect the Earth's surface through microwave remote sensing technology, such as a SAR dataset (a dataset collected by a satellite equipped with a C-band synthetic aperture radar).

[0060] Optionally, optical data refers to optical datasets that use remote sensing technology in the visible and near-infrared bands to detect the Earth's surface through optical signals and obtain information about the surface, such as surface vegetation cover, surface albedo, and surface temperature. For example, the MODIS dataset (a dataset collected by a satellite equipped with a high-spectral-resolution imaging spectrometer).

[0061] Optionally, the water level determination system can compensate for the shortcomings of a single remote sensing technology by simultaneously collecting both microwave and optical data. Microwave data has advantages in penetrating clouds and vegetation cover to detect surface information, while optical data shows high accuracy in monitoring vegetation changes and surface temperature. At the same time, microwave data can provide the water level determination system with information on moisture status such as soil moisture, while optical data can reflect variables such as surface cover type and surface temperature.

[0062] Optionally, the water level determination system combines microwave and optical data into multi-source remote sensing data, thereby providing a comprehensive dataset containing various surface information for the subsequent target inversion model. The combined use of microwave and optical data can more comprehensively and accurately reflect the hydrological status within the watershed to which the target reservoir belongs, overcoming the limitations of single remote sensing data, and thus providing richer and more comprehensive input data for the target inversion model.

[0063] In summary, through the above steps, the water level determination system, by referencing land data, makes the watershed boundary delineation of the target reservoir more accurate. This helps subsequent hydrological models focus on the hydrological evolution process of the area surrounding the target reservoir. The collection and integration of microwave and optical data not only overcomes the limitations of weather and geographical conditions on remote sensing monitoring technology, but also improves the comprehensiveness and accuracy of hydrological monitoring results through the complementarity of different types of data. This enhances the input characteristics of subsequent target inversion models, thereby enabling high-precision detection of the target reservoir's water level / water area even without runoff monitoring data.

[0064] In one optional embodiment, during the preprocessing of multi-source remote sensing data, the water level determination system first performs spatiotemporal alignment of the multi-source remote sensing data based on a preset resolution to obtain first remote sensing data with consistent resolution scale. The preset resolution includes at least a preset temporal resolution and a preset spatial resolution. Then, the water level determination system performs data cleaning and data correction on the first remote sensing data to obtain second remote sensing data corresponding to the target reservoir. Data cleaning removes optical data from the first remote sensing data where the cloud obscuration rate is greater than a preset obscuration rate, and data correction performs radiometric correction on microwave data in the first remote sensing data. Next, the water level determination system detects the data quality score corresponding to the second remote sensing data. The data quality score at least characterizes the radiometric accuracy and geometric accuracy of the data corresponding to the second remote sensing data. Subsequently, if the data quality score is less than a preset score, the water level determination system generates a reacquisition signal, which is used to reacquire the multi-source remote sensing data corresponding to the target reservoir. Finally, if the data quality score is greater than or equal to the preset score, the water level determination system uses the second remote sensing data as the target remote sensing data.

[0065] Optionally, the preset time resolution is ≤3 days, the preset spatial resolution is ≤100m, and the preset occlusion rate is <20%.

[0066] Optionally, preset temporal resolution and preset spatial resolution refer to the sampling time scale and sampling spatial scale of remote sensing data, which are key parameters to ensure the accuracy and consistency of data in time series and spatial analysis.

[0067] Optionally, the water level determination system performs spatiotemporal alignment of multi-source remote sensing data based on preset temporal and spatial resolutions. This matches and unifies different remote sensing data at both the sampling time and spatial scales, avoiding errors caused by temporal misalignment and spatial mismatch. It ensures that the input data of the target inversion model has accurate spatiotemporal resolution, thereby improving the target inversion model's ability to capture hydrological dynamics. Especially for short-term prediction tasks that require high-frequency and high-resolution data, the preset resolution setting ensures that the remote sensing data meets the task requirements.

[0068] Optionally, the water level determination system can effectively remove noise, outliers, and incomplete data from the first remote sensing data by performing data cleaning, especially removing optical data with cloud obscuration rates greater than a preset obscuration rate, thereby improving the accuracy and reliability of subsequent remote sensing data analysis results.

[0069] Optionally, the water level determination system performs data correction on the first remote sensing data to eliminate systematic and random deviations caused by factors such as instruments and atmospheric conditions during the acquisition process, especially performing radiometric correction on microwave data, thereby ensuring the correct physical meaning and magnitude of the subsequent model input data.

[0070] Optionally, the water level determination system improves the purity and accuracy of the input data for the subsequent target inversion model by cleaning and correcting the first remote sensing data, reducing the uncertainty in the model training and prediction process, thereby improving the reliability and prediction accuracy of the water level forecast results.

[0071] Optionally, the water level determination system can automatically filter out standard remote sensing data by detecting the data quality score corresponding to the second remote sensing data, avoiding prediction errors caused by data quality issues. When the data quality score is lower than the preset score standard, the water level determination system automatically triggers a recapture signal, instructing the system to re-acquire multi-source remote sensing data corresponding to the target reservoir, thus preventing the hydrological model from conducting hydrological monitoring based on low-quality remote sensing data.

[0072] Optionally, if the data quality score is greater than or equal to the preset score, the water level determination system uses the second remote sensing data as the target remote sensing data. This avoids the negative impact of low-quality data on the use and training process of the hydrological model, and also reduces the workload of system data processing and storage. This allows the water level determination system to make more efficient use of limited computing resources, thereby improving the efficiency of the overall reservoir water level prediction process.

[0073] In summary, by implementing the above steps, the water level determination system further strengthens the preprocessing and quality control of remote sensing data, ensuring the accuracy and applicability of the model input data. By performing spatiotemporal alignment, data cleaning, data correction, and data quality scoring on the collected multi-source remote sensing data, as well as a dynamic data resampling strategy based on the data quality score, an efficient and stable data preprocessing framework is formed, providing a high-quality data foundation for subsequent soil moisture content inversion and reservoir water level prediction.

[0074] In one optional embodiment, to accurately detect the soil moisture content data of the target reservoir, the water level determination system first inputs the target remote sensing data into the target inversion model. The target inversion model includes at least an inversion feature extraction layer and an inversion physical constraint layer. The target inversion model is a neural network model trained based on historical soil moisture content data and historical remote sensing data corresponding to P reservoirs, where P is a positive integer. Then, the water level determination system extracts features from the target remote sensing data through the inversion feature extraction layer, obtaining L surface features, where L is a positive integer. The L surface features are used to characterize at least the electromagnetic wave energy intensity, vegetation growth status, and surface temperature information of the watershed to which the target reservoir belongs. Next, the water level determination system determines the soil moisture content sequence of the target reservoir based on the L surface features. Subsequently, the water level determination system uses the inversion physical constraint layer to detect whether the soil moisture content sequence meets preset soil moisture constraint conditions. If the soil moisture content sequence meets the preset soil moisture constraint conditions, the water level determination system uses the soil moisture content sequence as the soil moisture content data of the target reservoir.

[0075] Optionally, the water level determination system can use observation data from ground observation points (such as soil moisture sensors) to calibrate the trained target inversion model. Furthermore, the inversion prediction results of the trained target inversion model need to be cross-validated, thereby providing a target inversion model with higher detection accuracy under low-cost deployment conditions.

[0076] Optionally, the water level determination system utilizes a target inversion model to detect soil moisture content based on target remote sensing data. The application of the target inversion model can achieve the following functions:

[0077] (1) Model generalization ability: The target inversion model trained based on historical soil moisture content and historical remote sensing data of multiple reservoirs has strong generalization ability, can adapt to the hydrological characteristics of different types of reservoirs, and improve the accuracy and applicability of soil moisture content inversion.

[0078] (2) Data-driven prediction: The target inversion model trained based on historical soil moisture content and historical remote sensing data of multiple reservoirs enables the trained model to automatically learn and capture the complex relationship between remote sensing signals and soil moisture content, thereby realizing the direct conversion from remote sensing data to soil moisture content and reducing the reliance on traditional ground monitoring.

[0079] Optionally, the water level determination system extracts features from the target remote sensing data through an inversion feature extraction layer. That is, the feature extraction layer integrates various remote sensing information corresponding to the target reservoir to obtain multi-dimensional surface features. The fusion of these multi-dimensional features helps the target inversion model to more comprehensively assess the hydrological state of the surface. At the same time, the system utilizes the deep learning automated feature extraction capability of the target inversion model to simplify the complexity of feature engineering, enabling the model to automatically identify the surface attributes that have the greatest influence on soil moisture content inversion.

[0080] Optionally, the water level determination system determines the soil moisture content sequence of the target reservoir based on L surface features, thereby converting abstract surface features into a soil moisture content sequence. This process not only provides soil moisture content information for a single point in the watershed to which the target reservoir belongs, but also captures the dynamic trend of soil moisture content changes, thus improving the accuracy of the subsequently predicted reservoir water level results. At the same time, the soil moisture content sequence inversion based on surface features can provide soil moisture information with higher spatial resolution than traditional ground observation points, further enhancing the monitoring accuracy of the hydrological status of the watershed to which the target reservoir belongs.

[0081] Optionally, the water level determination system detects whether the soil moisture content sequence meets the preset soil moisture constraint conditions by inverting the physical constraint layer. The preset soil moisture constraint conditions refer to the Richards equation (soil moisture movement equation). By constraining the soil moisture constraint conditions, a reasonable range of soil moisture content input to the model can be set based on the principles of hydrophysics. For example, the soil moisture content SM should be between 0 and the soil saturation moisture content SM_sat, that is, 0≤SM≤SM_sat must be satisfied. If it exceeds the range, it is truncated.

[0082] Optionally, the setting of the inversion physical constraint layer can ensure that the inversion results conform to the hydrophysical laws and avoid the distorted prediction results that may be generated by the data-driven model. The target inversion model only uses the soil moisture sequence as the soil moisture data of the target reservoir when the soil moisture sequence meets the preset soil moisture constraint conditions, which ensures the quality of the soil moisture data. That is, only the inversion results that meet the physical constraint conditions will be adopted, thereby improving the reliability of the final soil moisture data and providing accurate input basis for the subsequent target detection model to perform water level detection.

[0083] In summary, by performing the above steps, the water level determination system can accurately retrieve soil moisture content information of the reservoir basin using remote sensing technology and deep learning models, even without surface runoff monitoring data. This mechanism not only overcomes the challenge of data scarcity but also ensures the physical rationality of the retrieval results and the generalization ability of the model.

[0084] In one optional embodiment, to accurately detect the water level of the target reservoir, the water level determination system first inputs soil moisture content data and target precipitation data into a target detection model. The target detection model includes at least a long short-term memory (LSTM) network layer, an attention mechanism module, and a water level physical constraint layer. Then, the water level determination system extracts features from the soil moisture content data and target precipitation data through the LTM network layer to obtain the soil moisture content features and precipitation features corresponding to the target reservoir. Subsequently, the water level determination system uses the attention mechanism module to determine the water level based on the target reservoir's... Based on the climatic conditions of the watershed, the system determines the feature weights corresponding to soil moisture content and precipitation characteristics. Then, the water level determination system determines the rate of change of water volume in the target reservoir based on the soil moisture content characteristics, precipitation characteristics, and the corresponding feature weights. Next, the system checks whether the rate of change of water volume in the reservoir meets the preset water volume constraint conditions through the water level physical constraint layer. Finally, if the rate of change of water volume in the reservoir meets the preset water volume constraint conditions, the system determines the water level of the target reservoir based on the rate of change of water volume.

[0085] Optionally, Figure 2 This is an architecture diagram of an optional target detection model according to an embodiment of this application, such as... Figure 2 As shown, the target detection model includes an input layer, a long short-term memory network layer, an attention mechanism module, a water level physical constraint layer, and an output layer.

[0086] Optionally, the water level determination system comprehensively analyzes the soil moisture content data and precipitation data of the target reservoir based on the target detection model to predict the reservoir water level. By simultaneously using soil moisture content data and precipitation data as model input data, the water level determination system can more comprehensively describe hydrological conditions by utilizing the complementarity of the two types of data, thereby improving the accuracy of subsequent model water level prediction. The water level determination system uses a neural network model for prediction, which, compared with traditional statistical models, can learn more complex nonlinear data relationships and improve prediction capabilities, especially in cases where the data distribution is complex and the nonlinear relationship is obvious, thus completing the reservoir water level detection task more efficiently.

[0087] Optionally, Long Short-Term Memory (LSTM) is a special type of Recurrent Neural Network (RNN). This network solves the gradient vanishing / exploding problem of traditional RNNs by designing gating mechanisms, and can effectively capture long-term dependencies in sequence data. The core network structure of LSTM includes cell states and three gating mechanisms.

[0088] Optionally, the cell state is like an "information conveyor belt" that transmits long-term memory through linear operations. It is the core memory unit of LSTM. Its characteristic is that it maintains the stability of information flow through only a small number of linear interactions and can store key features across time steps (such as seasonal periodic patterns in hydrological sequences).

[0089] Optionally, the three gate control mechanisms are shown in List 1 below:

[0090] Table 1:

[0091]

[0092] Optionally, Sigmoid in the table above is a non-linear activation function, and Tanh is a hyperbolic tangent function.

[0093] Optionally, LSTM addresses the vanishing gradient problem of traditional recurrent neural networks (RNNs) through cell state and three gate control mechanisms, thereby effectively capturing long-term dependencies in hydrological sequences. The hydrological forecasting process using LSTM is as follows:

[0094] (1) Adaptation of gated unit structure to hydrological characteristics:

[0095] 1) Forget Gate: This gate uses a sigmoid activation function (outputting a value between 0 and 1) to dynamically determine whether to retain or discard historical hydrological information. For example, during the dry season, the forget gate weakens the weight of runoff data from previous flood seasons and focuses on current soil moisture and baseflow characteristics.

[0096] 2) Input Gate: Combining the candidate states generated by the tanh layer with the updated weights of the sigmoid layer, new information is selectively incorporated. In rainstorm events, the input gate strengthens the influence weight of real-time monitoring data such as rainfall intensity and soil moisture (usually with a weight value > 0.8).

[0097] 3) Cell State: Similar to a conveyor belt, this linear information flow achieves long-term memory of hydrological processes through the synergistic effect of forget gates and input gates. For example, in the simulation of glacial meltwater in the Qinghai-Tibet Plateau basin, the cell state can retain the cumulative impact of seasonal temperature changes on runoff for more than 6 months.

[0098] 4) Output Gate: Based on the current cell state and Sigmoid layer filtering, the output gate outputs features relevant to hydrological prediction. In water level forecasting, the output gate focuses on key parameters such as watershed lag time and confluence velocity.

[0099] Optionally, LSTM processes hourly / daytime-scale hydrological and meteorological sequences through a recurrent structure. The memory capacity of LSTM allows its Nash efficiency coefficient (NSE) to remain above 0.75 within a 72-hour forecast period, which is about 20% better than traditional models, thus ensuring the accuracy of the target detection model in detecting reservoir water levels.

[0100] As can be seen, the water level determination system extracts and processes features from soil moisture content data and target precipitation data through long short-term memory network layers. Compared with traditional RNNs, LSTM can better handle the gradient vanishing problem in sequential data, enabling the model to learn more stable feature representations from historical data and enhancing the stability and accuracy of prediction results.

[0101] Optionally, the water level determination system uses an attention mechanism module to determine the feature weights corresponding to soil moisture content and precipitation characteristics based on the climate conditions of the watershed to which the target reservoir belongs. The attention mechanism allows the model to automatically adjust the weights of soil moisture content and precipitation characteristics according to the climate conditions at different time points, thereby more accurately reflecting the dominant factors in the hydrological process. By optimizing the feature weights, the model can respond more flexibly to different hydrological scenarios and improve prediction accuracy. In particular, it can demonstrate more accurate hydrological prediction and processing performance when dealing with extreme climate events or complex water cycle conditions.

[0102] Optionally, the target detection model determines the rate of change of water volume in the target reservoir based on soil moisture content characteristics, precipitation characteristics, and the corresponding feature weights of soil moisture content characteristics and precipitation characteristics, which can achieve the following functions:

[0103] (1) Improve the accuracy of analysis: By combining the characteristic weights of soil moisture content and precipitation, the model can comprehensively analyze the impact of these two factors on the change of reservoir water volume, and provide a more accurate estimate of flow change for water level prediction.

[0104] (2) Integration of physical processes: By mapping soil moisture content characteristics and precipitation characteristics to the reservoir water volume change rate, the physical processes of the water cycle are essentially integrated into the model, thereby improving the physical rationality of the prediction results.

[0105] Optionally, the physical constraint condition in the water level physical constraint layer refers to the reservoir water balance equation ΔS=PER, where ΔS is the change in reservoir water volume, P is the precipitation, E is the evaporation, and R is the runoff. In this model, R is dynamically estimated from soil moisture data.

[0106] Optionally, the target detection model detects whether the rate of change of reservoir water volume meets the preset reservoir water volume constraint conditions through the water level physical constraint layer, which can achieve the following functions:

[0107] (1) Follow the physical laws: The water level physical constraint layer ensures that the predicted rate of change of reservoir water volume conforms to the basic physical laws of hydrology, such as the principle of water balance, which improves the credibility of the prediction results.

[0108] (2) Error correction mechanism: When the prediction results deviate from the range of physical feasibility, it can be identified and adjusted in time, avoiding unrealistic prediction results and enhancing the robustness of the model.

[0109] Optionally, the target detection model determines the water level of the target reservoir based on the reservoir water volume change rate, provided that the reservoir water volume change rate meets the preset reservoir water volume constraints. This ensures that the predicted reservoir water volume change rate is only used for water level prediction when it meets the physical constraints, thus guaranteeing the reliability and rationality of the final prediction results. Water level prediction based on the reservoir water volume change rate enables continuous monitoring of the reservoir water level and prediction of future water level change trends, providing information support for reservoir hydrological management decisions.

[0110] In summary, through the above steps, the water level determination system, based on the soil moisture content and precipitation data obtained by remote sensing technology, and by combining deep learning models with physical constraints, achieves high-precision prediction of water levels in reservoirs without runoff. Through dynamic feature weight allocation mechanism and physical constraint detection mechanism, it not only improves the flexibility and adaptability of the target detection model, but also improves the physical rationality and prediction accuracy of the prediction results.

[0111] In one optional embodiment, to train the target detection model, the water level determination system first collects historical water level data, historical soil moisture content data, and historical precipitation data from Q reservoirs without runoff, where Q is a positive integer. Then, the water level determination system divides the historical water level data, historical soil moisture content data, and historical precipitation data from the Q reservoirs into a training set and a validation set. Next, the water level determination system iteratively trains the initial model based on samples from the training set to obtain a training model. Then, the water level determination system determines the loss function value corresponding to the training model based on samples from the validation set. The loss function value is determined based on the water level prediction error, physical constraint loss, and the regularization term in the attention mechanism module of the training model. Subsequently, if the loss function value is less than or equal to a preset function value, the water level determination system uses the training model obtained from the last training as the target detection model.

[0112] Optionally, historical water level data, historical soil moisture content data, and historical precipitation data of Q reservoirs with no runoff are collected, with a time span of ≥3 years, and historical samples covering both dry and rainy seasons are included.

[0113] Optionally, the water level determination system uses historical water level data, historical soil moisture content data, and historical precipitation data from Q reservoirs without runoff as training samples for the target detection model, thus obtaining training samples covering different climatic conditions, geographical environments, and hydrological characteristics, thereby enhancing the diversity and representativeness of the model training data.

[0114] Optionally, the water level determination system divides the historical water level data, historical soil moisture content data, and historical precipitation data of Q reservoirs without runoff to obtain independent training and validation sets. The training set can independently evaluate the model's predictive ability, while the validation set is used to test the model's generalization ability and prediction error, avoid overfitting, and ensure the model's effectiveness on unknown data.

[0115] Optionally, the water level determination system iteratively trains the initial model based on samples in the training set to obtain a trained model. Through iterative training, the trained model can continuously adjust its parameters to minimize the prediction error on the training set and improve the prediction accuracy of the target detection model obtained in the final training. Based on extensive historical training data, the model can automatically learn and extract complex patterns in the hydrological process. In particular, the LSTM and attention mechanism modules can capture the weight distribution between long-term dependencies and time series features, thereby improving the accuracy of reservoir water level forecasting.

[0116] Optionally, the water level determination system determines the loss function value corresponding to the training model based on samples in the validation set. The loss function value is jointly determined by the water level prediction error of the training model, the physical constraint loss, and the regularization term in the attention mechanism module. It is a key indicator for evaluating model performance and optimization direction, and this mechanism has the following functions:

[0117] (1) Comprehensive performance index: The loss function not only considers the water level forecast error, but also incorporates physical constraint loss and regularization term to ensure that the model prediction results are both accurate and in line with hydrological and physical laws, thus avoiding unreasonable prediction results that may occur in pure data-driven prediction.

[0118] (2) Prevent overfitting: The role of regularization is to prevent the model from overfitting the training data during the training process and to improve the model's generalization ability on the validation set and even unknown data.

[0119] In summary, through the above steps, the water level determination system ensures that the trained target detection model can accurately predict the water level of the target reservoir based on the soil moisture content data and the target precipitation data, even without runoff monitoring data. During the model training process, physical constraints and regularization measures ensure the rationality and stability of the prediction results, thereby improving the water level detection accuracy and precision of the final trained target detection model.

[0120] In one optional embodiment, after determining the water level of the target reservoir based on soil moisture content data and target precipitation data, the water level determination system generates early warning information when the water level of the target reservoir is higher than or equal to a preset water level. The early warning information is used to issue flood warnings for the watershed to which the target reservoir belongs.

[0121] Optionally, a preset water level can be set, which is a threshold water level pre-set based on comprehensive factors such as the reservoir's capacity, historical water levels, and the flood tolerance of downstream areas. When the reservoir water level reaches or exceeds this threshold, the reservoir is considered to be at risk of flooding.

[0122] Optionally, the early warning information system will automatically generate an alarm message once the reservoir water level reaches or exceeds the preset water level. This message will be used to remind reservoir managers and downstream communities of the potential flood threat and to take timely flood control measures.

[0123] In summary, the water level determination system can identify flood risks in advance by comparing dynamically monitored reservoir water levels with preset water levels, providing reservoir managers with sufficient early warning time to formulate and implement emergency plans, and ensuring the safety of downstream communities and infrastructure. Through automated early warning information generation, it can help reservoir operators optimize water release scheduling strategies and avoid overshooting and safety risks caused by sudden floods.

[0124] In one alternative embodiment, Figure 3 A flowchart of an optional reservoir water level prediction method based on multi-source remote sensing data and a physically constrained neural network, according to an embodiment of this application, is shown below. Figure 3 The method includes the following steps:

[0125] Step 1: Acquisition and preprocessing of multi-source remote sensing data:

[0126] (1) Fusion of multi-source remote sensing data: The VV / VH (vertical transmission, vertical reception / vertical transmission, horizontal reception) polarization data collected by Sentinel-1 satellite in the SAR dataset is aligned with MODIS surface temperature to a unified spatiotemporal grid.

[0127] (2) Quality control: Optical data with cloud coverage > 20% are removed, and microwave data are radiatively corrected.

[0128] Output a spatiotemporally consistent remote sensing data cube (dimensions: time × space × band) to provide reliable input for the target inversion model.

[0129] Step 2: Remote sensing inversion of soil moisture content:

[0130] (1) Obtain the initial inversion model and load the calibration point data (soil moisture content data collected by soil moisture sensor) and historical remote sensing data corresponding to the reservoir.

[0131] (2) Based on the preprocessed historical remote sensing data, feature engineering is constructed. The feature engineering is represented in matrix form, i.e., X=[backscattering, NDVI, surface temperature], where NDVI stands for "Normalized Difference Vegetation Index".

[0132] (3) Based on feature engineering, the initial inversion model is iteratively trained to obtain the target inversion model. The key operations in the training process are as follows:

[0133] Add physical constraints: ensure that the inversion result satisfies 0≤SM≤SM_sat (saturated water content), and truncate if it exceeds the range.

[0134] Spatiotemporal interpolation: Kriging is used to fill in the missing data regions in the inversion results.

[0135] The model inversion results are daily series of soil moisture content in the watershed (unit: m³ / m³), with a detection accuracy of R²=0.75 (verified by field measurements), thus replacing runoff monitoring.

[0136] (4) Water volume inversion is performed using the target inversion model to obtain soil moisture content data corresponding to the target reservoir.

[0137] Step 3: Construct the initial PINN (Physics-Informed Neural Networks) model + attention mechanism module:

[0138] Optionally, the initial PINN model includes: an input layer, an LSTM layer, an attention mechanism module, a physical constraint layer, and an output layer.

[0139] Optionally, the input data for the attention mechanism module is: precipitation sequence. and soil moisture content sequence The output data of the attention mechanism module is: precipitation sequence weights. and soil moisture content sequence weights , in the sequence .

[0140] Optionally, the physical constraint layer is used to detect whether the rate of change of reservoir water volume detected by the PINN model conforms to the reservoir water balance equation.

[0141] Step 4: Model Training and Optimization

[0142] Optionally, the loss function during the initial PINN model iterative training process is: prediction error loss (MSE) + physical constraint loss (equation residual) + regularization term of the attention mechanism module.

[0143] Optionally, the equation residuals ,in, The rate of change in water level in the reservoir. For precipitation series, The soil moisture content sequence Let be the evaporation sequence, σ be the Sigmoid function, k∈[0.6,0.9] be the watershed soil type coefficient, and λ be the learnable weight parameter.

[0144] Training process:

[0145] a. Use historical water level data to divide the training set and validation set (the ratio is 8:2, that is, the data in the training set accounts for 80% of the time series length of the dataset corresponding to the full historical water level data).

[0146] b. Set the optimizer AdamW (lr=1e-4), and set the early stopping mechanism to terminate if the verification loss does not decrease for 5 consecutive rounds.

[0147] The trained PINN model achieved a Nash-Sutcliffe coefficient (a statistical metric used to evaluate the performance of hydrological models) of 0.85 on the validation set, with physical constraint loss accounting for less than 10%, ensuring physical rationality.

[0148] Step 5: Conduct real-time water level forecasting:

[0149] (1) Automatically acquire the latest precipitation data and soil moisture content data of the target reservoir every day.

[0150] (2) Input the trained PINN model and output the water level forecast for the next 7 days.

[0151] (3) Add a rationality check: If the predicted water level change rate is greater than 1.5 times the historical maximum value, the expert review process will be triggered.

[0152] Tests showed that the RMSE (Root Mean Square Error) prediction error of the trained PINN model is ≤0.15m, which is 32% lower than that of the traditional model.

[0153] As can be seen from the above, this application determines the soil moisture content data corresponding to the watershed of the reservoir with no runoff based on multi-source remote sensing data, and uses the soil moisture content data to replace the runoff monitoring data in traditional technology to predict the reservoir water level, thereby overcoming the technical bottleneck of scarce runoff monitoring data in the existing technology, and thus achieving the purpose of accurately predicting the reservoir water level.

[0154] Specifically, this application first unifies the spatiotemporal resolution of multi-source remote sensing data by preprocessing the pre-collected multi-source remote sensing data, thereby improving the accuracy of the acquired reservoir soil moisture content data and effectively filling the gap in reservoir water level monitoring in areas without monitoring stations. Then, this application determines the reservoir water level based on a two-layer factor of reservoir soil moisture content data and reservoir precipitation data, thereby achieving the technical effect of improving the timeliness and accuracy of reservoir water level detection results, and thus solving the technical problem of low accuracy in water level detection of reservoirs without runoff based on existing technologies.

[0155] Example 2

[0156] This application embodiment can also provide a reservoir water level determination device. It should be noted that the reservoir water level determination device of this application embodiment can be used to execute the reservoir water level determination method provided in this application embodiment. The reservoir water level determination device provided in this application embodiment will be described below.

[0157] According to an embodiment of this application, an apparatus for implementing the above-described method for determining reservoir water level is also provided. Figure 4 A schematic diagram of an optional reservoir water level determination device according to an embodiment of this application is shown below. Figure 4 As shown, the device includes: a data acquisition unit 401, a data preprocessing unit 402, a soil moisture content data determination unit 403, and a water level determination unit 404.

[0158] Optionally, the data acquisition unit 401 is used to acquire multi-source remote sensing data corresponding to the target reservoir, wherein the multi-source remote sensing data is used to characterize the surface information of the watershed to which the target reservoir belongs; the data preprocessing unit 402 is used to preprocess the multi-source remote sensing data to obtain the target remote sensing data corresponding to the target reservoir, wherein the preprocessing is at least used to unify the resolution scale corresponding to the multi-source remote sensing data; the soil moisture content data determination unit 403 is used to invert the soil moisture content data of the target reservoir based on the target remote sensing data; and the water level determination unit 404 is used to determine the water level of the target reservoir based on the soil moisture content data and the target precipitation data, wherein the target precipitation data is used to characterize the precipitation information of the watershed to which the target reservoir belongs.

[0159] In one optional embodiment, the data acquisition unit 401 includes: a watershed division subunit, a multi-source data acquisition subunit, and a multi-source data determination subunit.

[0160] Optionally, a watershed delineation subunit is used to delineate the watershed boundary of the target reservoir based on the land data corresponding to the target reservoir, thereby obtaining the watershed to which the target reservoir belongs. The land data includes at least the land type and land use of the watershed to which the target reservoir belongs. A multi-source data acquisition subunit is used to acquire microwave data and optical data corresponding to the watershed to which the target reservoir belongs. The microwave data is surface data obtained by detecting microwave signals, and the optical data is surface data obtained by detecting light wave signals. A multi-source data determination subunit is used to use the microwave data and optical data as multi-source remote sensing data corresponding to the target reservoir.

[0161] In one optional embodiment, the data preprocessing unit 402 includes: a spatiotemporal alignment subunit, a cleaning and correction subunit, a quality detection subunit, a multi-source remote sensing data recapture subunit, and a target remote sensing data inversion subunit.

[0162] Optionally, the spatiotemporal alignment subunit is used to perform spatiotemporal alignment of multi-source remote sensing data based on a preset resolution to obtain first remote sensing data with consistent resolution scale, wherein the preset resolution includes at least a preset temporal resolution and a preset spatial resolution; the cleaning and correction subunit is used to perform data cleaning and data correction on the first remote sensing data to obtain second remote sensing data corresponding to the target reservoir, wherein data cleaning is used to remove optical data in the first remote sensing data with a cloud obscuration rate greater than a preset obscuration rate, and data correction is used to perform radiometric correction on microwave data in the first remote sensing data; the quality detection subunit is used to detect the data quality score corresponding to the second remote sensing data, wherein the data quality score is used to characterize at least the radiometric accuracy and image geometric accuracy of the data corresponding to the second remote sensing data; the multi-source remote sensing data recapture subunit is used to generate a recapture signal when the data quality score is less than a preset score, wherein the recapture signal is used to reacquire multi-source remote sensing data corresponding to the target reservoir; and the target remote sensing data inversion subunit is used to use the second remote sensing data as the target remote sensing data when the data quality score is greater than or equal to the preset score.

[0163] In one optional embodiment, the soil moisture content data determination unit 403 includes: a first input subunit, a first extraction subunit, a soil moisture content sequence determination subunit, a first physical constraint detection subunit, and a soil moisture content data determination subunit.

[0164] Optionally, the first input subunit is used to input the target remote sensing data into the target inversion model, wherein the target inversion model includes at least an inversion feature extraction layer and an inversion physical constraint layer, and the target inversion model is a neural network model trained based on historical soil moisture content data and historical remote sensing data corresponding to P reservoirs, where P is a positive integer; the first extraction subunit is used to extract features from the target remote sensing data through the inversion feature extraction layer to obtain L surface features, where L is a positive integer, and the L surface features are used to characterize at least the electromagnetic wave energy intensity, vegetation growth status, and surface temperature information of the watershed to which the target reservoir belongs; the soil moisture content sequence determination subunit is used to determine the soil moisture content sequence of the target reservoir based on the L surface features; the first physical constraint detection subunit is used to detect whether the soil moisture content sequence meets the preset soil moisture constraint conditions through the inversion physical constraint layer; the soil moisture content data determination subunit is used to use the soil moisture content sequence as the soil moisture content data of the target reservoir if the soil moisture content sequence meets the preset soil moisture constraint conditions.

[0165] In one optional embodiment, the water level determination unit 404 includes: a second input subunit, a second extraction subunit, a feature weight determination subunit, a water volume change rate determination subunit, a second physical constraint detection subunit, and a water level determination subunit.

[0166] Optionally, the second input subunit is used to input soil moisture content data and target precipitation data into the target detection model, wherein the target detection model includes at least a long short-term memory network layer, an attention mechanism module, and a water level physical constraint layer; the second extraction subunit is used to extract features from the soil moisture content data and target precipitation data through the long short-term memory network layer to obtain the soil moisture content features and precipitation features corresponding to the target reservoir; the feature weight determination subunit is used to determine the soil weight based on the climatic conditions of the watershed to which the target reservoir belongs through the attention mechanism module. The system includes: a feature weighting system for soil moisture content characteristics and precipitation characteristics; a water change rate determination subunit, used to determine the water change rate of the target reservoir based on soil moisture content characteristics, precipitation characteristics, and the feature weights corresponding to soil moisture content characteristics and precipitation characteristics; a second physical constraint detection subunit, used to detect whether the water change rate of the reservoir meets the preset reservoir water constraint conditions through the water level physical constraint layer; and a water level determination subunit, used to determine the water level of the target reservoir based on the water change rate of the reservoir if the water change rate meets the preset reservoir water constraint conditions.

[0167] In one optional embodiment, the reservoir water level determination device further includes: a historical data acquisition unit, a historical data partitioning unit, an iterative training unit, a loss function value determination unit, and a target detection model determination unit.

[0168] Optionally, the historical data acquisition unit is used to collect historical water level data, historical soil moisture content data, and historical precipitation data of Q reservoirs without runoff, where Q is a positive integer; the historical data partitioning unit is used to partition the historical water level data, historical soil moisture content data, and historical precipitation data of the Q reservoirs without runoff into training sets and validation sets; the iterative training unit is used to iteratively train the initial model based on samples in the training set to obtain a training model; the loss function value determination unit is used to determine the loss function value corresponding to the training model based on samples in the validation set, wherein the loss function value is determined based on the water level prediction error, physical constraint loss, and regularization term in the attention mechanism module of the training model; and the target detection model determination unit is used to use the training model obtained from the last training as the target detection model if the loss function value is less than or equal to a preset function value.

[0169] In one optional embodiment, the reservoir water level determination device further includes an early warning information generation unit.

[0170] Optionally, the early warning information generation unit is used to generate early warning information when the water level of the target reservoir is higher than or equal to a preset water level after determining the water level of the target reservoir based on the soil moisture content data and the target precipitation data. The early warning information is used to issue flood warnings for the watershed to which the target reservoir belongs.

[0171] As can be seen from the above, this application determines the soil moisture content data corresponding to the watershed of the reservoir with no runoff based on multi-source remote sensing data, and uses the soil moisture content data to replace the runoff monitoring data in traditional technology to predict the reservoir water level, thereby overcoming the technical bottleneck of scarce runoff monitoring data in the existing technology, and thus achieving the purpose of accurately predicting the reservoir water level.

[0172] Specifically, this application first unifies the spatiotemporal resolution of multi-source remote sensing data by preprocessing the pre-collected multi-source remote sensing data, thereby improving the accuracy of the acquired reservoir soil moisture content data and effectively filling the gap in reservoir water level monitoring in areas without monitoring stations. Then, this application determines the reservoir water level based on a two-layer factor of reservoir soil moisture content data and reservoir precipitation data, thereby achieving the technical effect of improving the timeliness and accuracy of reservoir water level detection results, and thus solving the technical problem of low accuracy in water level detection of reservoirs without runoff based on existing technologies.

[0173] It should be noted that the data acquisition unit 401, data preprocessing unit 402, soil moisture content data determination unit 403, and water level determination unit 404 mentioned above correspond to steps S101 to S104 in the method embodiment. The instances and application scenarios implemented by the above units and corresponding steps are the same, but are not limited to the content disclosed in the above embodiment.

[0174] Example 3

[0175] Embodiments of this application can also provide an electronic device. Figure 5 This is a structural block diagram of an electronic device according to an embodiment of this application, such as... Figure 5 As shown, the electronic device includes: one or more ( Figure 5 (Only one is shown) processor 502, memory 504, memory controller, and peripheral interface, wherein the peripheral interface is connected to the radio frequency module, audio module and display.

[0176] The memory can be used to store software programs and modules, such as the program instructions / modules corresponding to the methods and devices in the embodiments of this application. The processor executes various functional applications and data processing by running the software programs and modules stored in the memory, thereby realizing the above-mentioned method for determining the reservoir water level.

[0177] The memory may include high-speed random access memory (RAM), and may also include non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some instances, the memory may further include memory remotely located relative to the processor, which can be connected to the terminal via a network. Examples of such networks include, but are not limited to, the Internet, intranets, local area networks (LANs), mobile communication networks, and combinations thereof.

[0178] The processor can access the information and application programs stored in the memory via a transmission device to execute the following steps: acquiring multi-source remote sensing data corresponding to the target reservoir, wherein the multi-source remote sensing data is used to characterize the surface information of the watershed to which the target reservoir belongs; preprocessing the multi-source remote sensing data to obtain target remote sensing data corresponding to the target reservoir, wherein the preprocessing is at least used to unify the resolution scale corresponding to the multi-source remote sensing data; inverting the soil moisture content data of the target reservoir based on the target remote sensing data; determining the water level of the target reservoir based on the soil moisture content data and the target precipitation data, wherein the target precipitation data is used to characterize the precipitation information of the watershed to which the target reservoir belongs.

[0179] This application provides a scheme for determining reservoir water levels. Based on multi-source remote sensing data, this application determines the soil moisture content data corresponding to the watershed of a reservoir without runoff. This soil moisture content data replaces the runoff monitoring data used in traditional techniques to predict reservoir water levels, thus overcoming the technical bottleneck of scarce runoff monitoring data in existing technologies and achieving the goal of accurate reservoir water level forecasting.

[0180] Specifically, this application first unifies the spatiotemporal resolution of multi-source remote sensing data by preprocessing the pre-collected multi-source remote sensing data, thereby improving the accuracy of the acquired reservoir soil moisture content data and effectively filling the gap in reservoir water level monitoring in areas without monitoring stations. Then, this application determines the reservoir water level based on a two-layer factor of reservoir soil moisture content data and reservoir precipitation data, thereby achieving the technical effect of improving the timeliness and accuracy of reservoir water level detection results, and thus solving the technical problem of low accuracy in water level detection of reservoirs without runoff based on existing technologies.

[0181] Those skilled in the art will understand that Figure 5 The structure shown is for illustrative purposes only. Electronic devices can also be smartphones, tablets, PDAs, mobile internet devices, PADs, and other terminal devices. Figure 5 This does not limit the structure of the aforementioned electronic device. For example, electronic devices may also include components that are more... Figure 5 The more or fewer components shown (such as network interfaces, display devices, etc.), or having the same Figure 5 The different configurations shown.

[0182] Those skilled in the art will understand that all or part of the steps in the various methods of the above embodiments can be implemented by a program instructing the hardware related to the terminal device. The program can be stored in a computer-readable storage medium, which may include: flash drive, read-only memory (ROM), random access memory (RAM), disk or optical disk, etc.

[0183] Example 4

[0184] Embodiments of this application may also provide a storage medium.

[0185] Optionally, in this embodiment of the application, the storage medium can be used to store the program code executed by the method for determining the reservoir water level provided in the above method embodiment.

[0186] Optionally, in this embodiment, the storage medium may be located in any computer terminal in a group of computer terminals in a computer network, or in any mobile terminal in a group of mobile terminals.

[0187] This application also provides a computer program product, which, when executed on a data processing device, is suitable for performing the steps of a method for determining the water level of a reservoir.

[0188] The sequence numbers of the embodiments in this application are for descriptive purposes only and do not represent the superiority or inferiority of the embodiments.

[0189] In the above embodiments of this application, the descriptions of each embodiment have different focuses. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions of other embodiments.

[0190] In the several embodiments provided in this application, it should be understood that the disclosed technical content can be implemented in other ways. The device embodiments described above are merely illustrative; for example, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the displayed or discussed mutual coupling, direct coupling, or communication connection may be through some interfaces; the indirect coupling or communication connection between units or modules may be electrical or other forms.

[0191] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0192] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0193] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as a USB flash drive, read-only memory (ROM), random access memory (RAM), portable hard drive, magnetic disk, or optical disk.

[0194] The above description is only a preferred embodiment of this application. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the principle of this application, and these improvements and modifications should also be considered within the scope of protection of this application.

Claims

1. A method for determining the water level of a reservoir, characterized in that, include: Collect multi-source remote sensing data corresponding to the target reservoir, wherein the multi-source remote sensing data is used to characterize the surface information of the watershed to which the target reservoir belongs; The multi-source remote sensing data is preprocessed to obtain the target remote sensing data corresponding to the target reservoir, wherein the preprocessing is at least used to unify the resolution scale corresponding to the multi-source remote sensing data; Soil moisture content data of the target reservoir were obtained by inversion based on the remote sensing data of the target. The water level of the target reservoir is determined based on the soil moisture content data and the target precipitation data, wherein the target precipitation data is used to characterize the precipitation information of the watershed to which the target reservoir belongs.

2. The method for determining reservoir water level according to claim 1, characterized in that, Collect multi-source remote sensing data corresponding to the target reservoir, including: Based on the land data corresponding to the target reservoir, the watershed boundary of the target reservoir is delineated to obtain the watershed to which the target reservoir belongs. The land data includes at least the land type and land use of the watershed to which the target reservoir belongs. Microwave data and optical data corresponding to the watershed to which the target reservoir belongs are collected, wherein the microwave data is surface data obtained by detecting microwave signals, and the optical data is surface data obtained by detecting light wave signals; The microwave data and the optical data are used as the multi-source remote sensing data corresponding to the target reservoir.

3. The method for determining reservoir water level according to claim 1, characterized in that, The multi-source remote sensing data is preprocessed to obtain the target remote sensing data corresponding to the target reservoir, including: The multi-source remote sensing data is spatiotemporally aligned based on a preset resolution to obtain first remote sensing data with the same resolution scale, wherein the preset resolution includes at least a preset temporal resolution and a preset spatial resolution. The first remote sensing data is cleaned and corrected to obtain the second remote sensing data corresponding to the target reservoir. The data cleaning is used to remove optical data in the first remote sensing data where the cloud obscuration rate is greater than a preset obscuration rate. The data correction is used to perform radiometric correction on the microwave data in the first remote sensing data. The data quality score corresponding to the second remote sensing data is detected, wherein the data quality score is used at least to characterize the data radiometric accuracy and image geometric accuracy corresponding to the second remote sensing data; If the data quality score is less than the preset score, a recapture signal is generated, wherein the recapture signal is used to reacquire the multi-source remote sensing data corresponding to the target reservoir; If the data quality score is greater than or equal to the preset score, the second remote sensing data will be used as the target remote sensing data.

4. The method for determining reservoir water level according to claim 1, characterized in that, Soil moisture content data of the target reservoir is obtained by inversion based on the remote sensing data of the target, including: The target remote sensing data is input into the target inversion model, wherein the target inversion model includes at least an inversion feature extraction layer and an inversion physical constraint layer, and the target inversion model is a neural network model trained based on historical soil moisture content data and historical remote sensing data corresponding to P reservoirs, where P is a positive integer; The target remote sensing data is subjected to feature extraction through the inversion feature extraction layer to obtain L surface features, where L is a positive integer. The L surface features are used to characterize at least the electromagnetic wave energy intensity, vegetation growth status and surface temperature information of the watershed to which the target reservoir belongs. The soil moisture content sequence of the target reservoir is determined based on the L surface features; The inversion physical constraint layer is used to detect whether the soil moisture content sequence meets the preset soil moisture constraint conditions. If the soil moisture content sequence meets the preset soil moisture constraint conditions, the soil moisture content sequence shall be used as the soil moisture content data of the target reservoir.

5. The method for determining reservoir water level according to claim 1, characterized in that, Determining the water level of the target reservoir based on soil moisture content data and target precipitation data includes: The soil moisture content data and the target precipitation data are input into the target detection model, wherein the target detection model includes at least a long short-term memory network layer, an attention mechanism module, and a water level physical constraint layer; The soil moisture content data and the target precipitation data are extracted by the long short-term memory network layer to obtain the soil moisture content characteristics and precipitation characteristics corresponding to the target reservoir. The attention mechanism module determines the feature weights corresponding to the soil moisture content characteristics and the precipitation characteristics based on the climate conditions of the watershed to which the target reservoir belongs. Based on the soil moisture content characteristics, the precipitation characteristics, and the feature weights corresponding to the soil moisture content characteristics and the precipitation characteristics, the rate of change of water volume in the target reservoir is determined; The water level physical constraint layer is used to detect whether the rate of change of the reservoir water volume meets the preset reservoir water volume constraint conditions. If the rate of change of water volume in the reservoir meets the preset water volume constraint, the water level of the target reservoir is determined based on the rate of change of water volume in the reservoir.

6. The method for determining reservoir water level according to claim 5, characterized in that, The training steps of the object detection model include: Historical water level data, historical soil moisture content data, and historical precipitation data were collected from Q reservoirs with no runoff, where Q is a positive integer; The historical water level data, historical soil moisture content data, and historical precipitation data of the Q reservoirs without runoff are divided into training set and validation set; The initial model is iteratively trained based on the samples in the training set to obtain the trained model; The loss function value corresponding to the training model is determined based on the samples in the validation set, wherein the loss function value is determined based on the water level prediction error, physical constraint loss and regularization term in the attention mechanism module of the training model; If the loss function value is less than or equal to a preset function value, the training model obtained from the last training session will be used as the target detection model.

7. The method for determining reservoir water level according to claim 5, characterized in that, After determining the water level of the target reservoir based on soil moisture content data and target precipitation data, the method for determining the reservoir water level further includes: When the water level of the target reservoir is higher than or equal to a preset water level, an early warning message is generated, wherein the early warning message is used to issue a flood warning for the watershed to which the target reservoir belongs.

8. A device for determining the water level of a reservoir, characterized in that, include: The data acquisition unit is used to acquire multi-source remote sensing data corresponding to the target reservoir, wherein the multi-source remote sensing data is used to characterize the surface information of the watershed to which the target reservoir belongs; A data preprocessing unit is used to preprocess the multi-source remote sensing data to obtain target remote sensing data corresponding to the target reservoir, wherein the preprocessing is at least used to unify the resolution scale corresponding to the multi-source remote sensing data. The soil moisture content data determination unit is used to invert the soil moisture content data of the target reservoir based on the target remote sensing data; A water level determination unit is used to determine the water level of the target reservoir based on soil moisture content data and target precipitation data, wherein the target precipitation data is used to characterize the precipitation information of the watershed to which the target reservoir belongs.

9. A computer program product, characterized in that, The computer program product includes a computer program, wherein, when the computer program is executed, it controls the computer program product to perform the method for determining the reservoir water level as described in any one of claims 1 to 7.

10. An electronic device, characterized in that, It includes one or more processors and a memory, the memory being used to store one or more programs, wherein when the one or more programs are executed by the one or more processors, the one or more processors cause the one or more processors to implement the method for determining the reservoir water level as described in any one of claims 1 to 7.

Images