Reservoir real-time flood forecasting method and system based on multi-source data fusion technology

Abstract

The application discloses a reservoir real-time flood forecasting method and system based on a multi-source data fusion technology, belongs to the technical field of reservoir flood forecasting, and comprises the following steps: acquiring multi-source water level monitoring data and performing preprocessing; performing data fusion on the processed multi-source water level monitoring data, and calculating actual water levels after data fusion; deriving inflow into a reservoir according to the actual water levels and real-time discharge flow data; inputting the inflow into a hydrological forecasting model for dynamic assimilation to generate a flood forecasting result. Through high-precision multi-source data fusion and dynamic assimilation technology, combined with deep learning, video image recognition technology and hydrological model assimilation and other frontier technologies, the application significantly improves the accuracy, timeliness and reliability of reservoir flood warning, significantly improves the real-time flood control capacity of the reservoir, and provides reliable technical support for flood control and disaster reduction.

Description

Technical Field

[0001] This application belongs to the field of reservoir flood forecasting technology, and specifically relates to a method and system for real-time reservoir flood forecasting based on multi-source data fusion technology. Background Technology

[0002] Real-time flood forecasting of reservoirs is a crucial link in flood control and disaster reduction, and is of great significance to safeguarding people's lives and property. Traditional flood forecasting methods mainly rely on hydrological models, combined with historical precipitation and runoff data, and a small amount of real-time monitoring data, to predict flood evolution through numerical simulation. However, under rapidly changing extreme weather conditions, these methods are limited by the accuracy of real-time data and the uncertainty of the models, and there is still considerable room for improvement in prediction accuracy. In terms of real-time monitoring, due to factors such as topography, water flow patterns, and difficulties in instrument deployment, it is often difficult to accurately monitor the inflow into the reservoir. Instead, the inflow is estimated based on the water balance method, using comprehensive information such as reservoir water level changes and outflow. Traditional water level monitoring technologies mainly include float-type water level gauges, pressure water level gauges, and ultrasonic water level gauges. These technologies obtain reservoir water level data through direct or indirect physical measurement methods. Their advantage is that the equipment is mature and applicable to most reservoir operation scenarios. However, under extreme hydrological conditions such as floods, water levels fluctuate dramatically, and monitoring equipment may be affected by floating debris or waves, leading to inaccurate monitoring data. Furthermore, some equipment is susceptible to wear and corrosion during long-term operation, affecting its lifespan and accuracy, thus impacting the accuracy of inflow estimation. With the development of UAV remote sensing technology, high-resolution water surface images acquired by UAVs, combined with water level inversion technology, can achieve efficient monitoring of reservoir water levels. This method has advantages such as wide coverage and fast response, showing significant advantages, especially in areas inaccessible or restricted by traditional equipment. However, the timeliness, accuracy, and environmental conditions of remote sensing data remain challenges. On the other hand, water level monitoring based on high-point video image recognition technology is gradually becoming an emerging method, with computer vision technology capable of extracting water level information from video images. Although this method provides a new perspective on water level monitoring, its accuracy and real-time performance still have room for improvement under different hydrological conditions and are usually limited by factors such as video quality and viewing angle. Traditional hydrological forecasting models typically rely on historical precipitation, evaporation, and infiltration data, combined with watershed hydrological characteristics, for simulation and prediction. However, model accuracy is limited by the availability of real-time data. When data is incomplete, model prediction errors are significant, especially due to a lack of sensitivity to real-time reservoir dynamics (such as changes in inflow), making it difficult to quickly adjust and correct the model's state. In recent years, with the rapid development of remote sensing, computer vision, and hydrological models, reservoir flood forecasting methods have gradually shifted from single physical modeling to multi-source data fusion and intelligent approaches. However, most existing technologies remain at the stage of simple data fusion, lacking targeted data processing and intelligent optimization methods. They cannot dynamically adjust weights and processing strategies based on changes in real-time monitoring data, thus limiting their application effectiveness in complex hydrological environments. Summary of the Invention

[0003] To address the aforementioned problems, this invention provides a method and system for real-time flood forecasting of reservoirs based on multi-source data fusion technology, thereby solving the problems of insufficient accuracy and limited dynamic adjustment capability of traditional methods under complex conditions.

[0004] A real-time flood forecasting method for reservoirs based on multi-source data fusion technology includes: Acquire multi-source water level monitoring data and perform preprocessing; The processed multi-source water level monitoring data are fused, and the actual water level after data fusion is calculated. The inflow rate is calculated based on the actual water level and real-time outflow data. The inflow is input into the hydrological forecasting model for dynamic assimilation to generate flood forecast results.

[0005] According to a specific embodiment of the present invention, the multi-source water level monitoring data includes high-point video monitoring data, UAV remote sensing data, and water level sensor monitoring data.

[0006] According to a specific embodiment of the present invention, acquiring and preprocessing multi-source water level monitoring data further includes: Acquire high-point video monitoring data and perform image denoising, edge detection, and marker localization; Acquire UAV remote sensing data and perform image registration, water surface extraction, and water level inversion processing; The system acquires water level sensor monitoring data, performs time series smoothing and error correction, and filters outliers.

[0007] According to a specific embodiment of the present invention, acquiring UAV remote sensing data and performing image registration, water surface extraction, and water level inversion processing further includes: Geometric registration of water surface areas in remote sensing images taken at different times and angles; Water surface areas are extracted from remote sensing images, water level height is retrieved through image analysis, and water level changes are estimated by combining historical topographic data and reservoir capacity curves.

[0008] According to a specific embodiment of the present invention, the process of fusing processed multi-source water level monitoring data and calculating the actual water level after data fusion further includes: Construct a multi-source data quality assessment system and dynamically calculate the first confidence weight of each data source; The first confidence weight is optimized by combining the entropy method and expert experience weight to obtain the second confidence weight; The processed multi-source water level monitoring data are fused based on the second confidence weight to calculate the actual water level after the fusion of multiple data sources.

[0009] According to a specific embodiment of the present invention, constructing a multi-source data quality assessment system and dynamically calculating the first confidence weight of each data source further includes: Calculate the residual standard deviation of each data source and the fusion result over N time periods. The calculation formula is as follows:

[0010] In the formula, Let be the observation value of the i-th data source at time k. The result of the fusion; Calculate the degree of deviation between each data source and the median of the multi-source data at the current time. The calculation formula is as follows:

[0011] In the formula, Let be the observation value of the i-th data source at time t. The median value of the multi-source data at time t; Based on residual standard deviation and degree of deviation Calculate the first confidence weight for each data source. The calculation formula is as follows:

[0012] In the formula, and To adjust the parameters, M represents the data source.

[0013] According to a specific embodiment of the present invention, the second confidence weight is further optimized by combining the entropy method and expert experience weight, including: A sliding window mechanism is used to dynamically update the weight allocation strategy. Combining the entropy method and expert experience weights, the first confidence weight is optimized through game theory to obtain the second confidence weight. The calculation formula is as follows:

[0014] In the formula, For objective weighting, Subjective weighting, This is the adjustment coefficient.

[0015] According to a specific embodiment of the present invention, the formula for calculating the actual water level is as follows:

[0016] In the formula, z is the actual water level value after multi-source data fusion; These are the water level values ​​monitored by each data source; The second confidence weights for each data source.

[0017] According to a specific embodiment of the present invention, deriving the inflow rate based on the actual water level and real-time outflow data further includes: The change in reservoir capacity is calculated based on the water level-capacity curve, and the formula is as follows:

[0018]

[0019] Where V is the reservoir capacity and z is the actual water level. The relationship between water level and reservoir capacity is shown in the curve. This represents the change in reservoir capacity within the scheduling period t. The real-time outflow rate is obtained from the flow meter, and the real-time inflow rate is calculated and output using the water balance formula. The calculation formula is as follows:

[0020] in, This refers to the inbound flow rate; t represents the outbound flow rate, and t represents the scheduling period.

[0021] According to a specific embodiment of the present invention, the process of inputting the inflow into a hydrological forecasting model for dynamic assimilation to generate flood forecasting results further includes: The hydrological forecasting model is initialized based on historical water level monitoring data, which includes historical water level, flow rate and rainfall data. Real-time acquisition of water level monitoring data and outflow rate, and calculation of inflow rate based on water balance formula; The calculated inflow is used as assimilated information and input into the hydrological forecasting model to generate flood prediction values; Based on the error between the flood forecast and the actual value, the assimilation parameters in the model are fed back in real time and dynamically updated. The hydrological forecasting model is dynamically revised based on the updated assimilation parameters to generate optimal flood forecasting results.

[0022] A real-time flood forecasting system for reservoirs based on multi-source data fusion technology includes: The data acquisition and processing module is used to acquire and preprocess multi-source water level monitoring data; The data fusion and calculation module is used to fuse the processed multi-source water level monitoring data and calculate the actual water level after data fusion. The flow rate calculation module is used to calculate the inflow rate based on the actual water level and real-time outflow data. The data assimilation and prediction module is used to dynamically assimilate the inflow into the hydrological forecasting model and generate the optimal flood forecast results.

[0023] According to a specific embodiment of the present invention, the data fusion and calculation module further includes: The first calculation module is used to construct a multi-source data quality assessment system and dynamically calculate the first confidence weight of each data source. The second calculation module is used to optimize the first confidence weight by combining the entropy method and expert experience weight to obtain the second confidence weight; The third calculation module is used to perform data fusion on the processed multi-source water level monitoring data based on the second confidence weight, and calculate the actual water level after the fusion of multiple data sources.

[0024] According to a specific embodiment of the present invention, the traffic back-calculation module further includes: The reservoir capacity calculation module is used to calculate the change in reservoir capacity based on the water level-capacity curve. The inflow calculation module is used to obtain the real-time outflow from the flow meter, calculate and output the real-time inflow using the water balance formula.

[0025] According to a specific embodiment of the present invention, the data assimilation and prediction module further includes: The model initialization module is used to initialize the hydrological forecasting model based on historical water level monitoring data, which includes historical water level, flow rate and rainfall data. The calculation module is used to acquire water level monitoring data and outflow in real time, and calculate inflow according to the water balance formula. The data assimilation module is used to input the calculated inflow as assimilation information into the hydrological forecasting model to generate flood prediction values; The parameter update module is used to provide real-time feedback and dynamically update the assimilation parameters in the model based on the error between the flood forecast and the actual values. The prediction module is used to dynamically revise the hydrological forecasting model based on updated assimilation parameters to generate optimal flood forecast results.

[0026] An electronic device includes a processor and a memory, wherein a computer program is stored in the memory, and the computer program is loaded and executed by the processor to implement the above-described method for real-time flood forecasting of reservoirs based on multi-source data fusion technology.

[0027] A computer-readable storage medium storing a computer program, which is loaded and executed by a processor to implement the above-described real-time flood forecasting method for reservoirs based on multi-source data fusion technology.

[0028] Compared with the prior art, this application has the following advantages: 1) Improve the reliability and accuracy of water level monitoring data. This invention integrates high-point video image recognition technology, UAV remote sensing data, and traditional water level sensor data to provide high-precision water level monitoring data in complex reservoir environments and under special conditions such as flood season. By calculating objective weights using the entropy method and combining expert experience, a game theory-based combinatorial optimization formula is constructed, thereby achieving effective fusion of multi-source water level monitoring data. This effectively avoids errors that may result from a single data source and can reflect dynamic changes in water levels in real time and accurately, thus improving the reliability and accuracy of water level monitoring and further enhancing the accuracy of reservoir flood forecasting.

[0029] 2) Accurately calculate inbound flow rate to improve flow monitoring accuracy. This invention achieves accurate calculation of inflow by precisely acquiring water level and outflow data and combining them with the water balance formula to inversely deduce inflow. This technology can provide inflow data in real time and accurately, serving as an important boundary condition for hydrological forecasting models. It effectively reduces forecast deviations caused by inflow errors, improving the accuracy of flow monitoring and the reliability of reservoir scheduling.

[0030] 3) Real-time assimilation of inbound traffic reduces model error. This invention uses real-time inflow data as assimilated into a hydrological forecasting model, enabling dynamic adjustment of the model's state and reducing errors. Specifically, the Kalman filter algorithm dynamically optimizes model weights, allowing the hydrological forecasting model to make accurate corrections based on real-time data in rapidly changing flood scenarios. This technology effectively improves the accuracy and real-time performance of hydrological models, further enhancing the reliability of flood forecasts and strengthening the real-time flood control capabilities and response speed of reservoirs.

[0031] 4) Improve the timeliness and accuracy of flood warnings. Other features and advantages of this application will be set forth in the description which follows, and will be apparent in part from the description, or may be learned by practicing the application. The objectives and other advantages of this application may be realized and obtained by means of the structures pointed out in the description, claims and drawings. Attached Figure Description

[0032] To more clearly illustrate the technical solutions in the embodiments of this disclosure or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this disclosure. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0033] Figure 1 This is a flowchart of a real-time flood forecasting method for reservoirs based on multi-source data fusion technology, provided according to an embodiment of the present invention.

[0034] Figure 2 This is a flowchart of a method for acquiring and preprocessing multi-source water level monitoring data according to an embodiment of the present invention.

[0035] Figure 3 This is a flowchart of a method for preprocessing remote sensing data from unmanned aerial vehicles (UAVs) according to an embodiment of the present invention.

[0036] Figure 4 This is a flowchart of a method for data fusion of multi-source water level monitoring data according to an embodiment of the present invention.

[0037] Figure 5 This is a flowchart of a method for constructing a multi-source data quality assessment system and dynamically calculating the first confidence weight of each data source, according to an embodiment of the present invention.

[0038] Figure 6 This is a flowchart of a method for deriving the inflow rate based on the actual water level and real-time outflow data according to an embodiment of the present invention.

[0039] Figure 7 This is a flowchart of a method for dynamically assimilating inflow into a hydrological forecasting model according to an embodiment of the present invention.

[0040] Figure 8 This is a structural diagram of a real-time flood forecasting system for reservoirs based on multi-source data fusion technology, provided according to an embodiment of the present invention.

[0041] Figure 9 This is a structural diagram of a data acquisition and processing module provided according to an embodiment of the present invention.

[0042] Figure 10 This is a structural diagram of a flow inversion module provided according to an embodiment of the present invention.

[0043] Figure 11 This is a structural diagram of a data assimilation and prediction module provided according to an embodiment of the present invention.

[0044] Figure 12 This is a schematic diagram of a computer device structure according to an embodiment of the present invention.

[0045] Figure label: 01-Data Acquisition and Processing Module; 02-Data Fusion and Calculation Module; 03-Flow Backpropagation Module; 04-Data Assimilation and Prediction Module; 011 - First calculation module; 012 - Second calculation module; 013 - Third calculation module; 031-Reservoir capacity calculation module; 032-Inflow rate calculation module; 041-Model initialization module; 042-Calculation module; 043-Data assimilation module; 044-Parameter update module; 045-Prediction module. Detailed Implementation

[0046] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0047] Example 1 Additional aspects and advantages of embodiments of the invention will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of embodiments of the invention. Figures 1-7 This invention provides a method for real-time flood forecasting of reservoirs based on multi-source data fusion technology, including: S1: Acquire multi-source water level monitoring data and perform preprocessing.

[0048] S2: Perform data fusion on the processed multi-source water level monitoring data and calculate the actual water level after data fusion.

[0049] S3: Estimate the inflow rate based on the actual water level and real-time outflow data.

[0050] S4: Input the inflow into the hydrological forecasting model for dynamic assimilation to generate flood forecast results.

[0051] Specifically, in step S1, multi-source water level monitoring data is acquired and preprocessed. This multi-source water level monitoring data includes high-point video monitoring data, UAV remote sensing data, and water level sensor monitoring data. In this embodiment of the invention, high-definition cameras installed at high points upstream or around the reservoir acquire real-time video images of the reservoir surface, obtaining high-point video monitoring data. The video equipment should have high-definition resolution and stable recording capabilities to ensure clear capture of water level markers (such as dam scale lines, buoys, or other water level reference marks). High-precision remote sensing equipment (such as lidar and optical sensors) mounted on a UAV periodically or as needed acquires multi-dimensional water level information of the reservoir area, obtaining UAV remote sensing data. Water level data acquired through remote sensing technology can not only cover areas inaccessible to traditional sensors but also provide broader and more detailed information on water level changes through remote sensing imagery. Real-time water level monitoring is performed using existing water level gauges, pressure sensors, and other equipment, obtaining water level sensor monitoring data. Traditional water level gauges (such as float-type and pressure-type water level gauges) monitor water levels in real-time at different locations in the reservoir, outputting accurate water level readings. These sensors can provide high-frequency and high-accuracy real-time monitoring data for water level.

[0052] Specifically, step S1, acquiring and preprocessing multi-source water level monitoring data, further includes: S11: Acquire high-point video monitoring data and perform image denoising, edge detection, and marker localization.

[0053] In this embodiment of the invention, the acquired real-time video frames are transmitted to an image processing system. Image processing algorithms (such as Convolutional Neural Networks, CNNs) are used to remove noise from the images, ensuring that the edges of the water surface area and water level markings are clearly visible. Edge detection algorithms (such as the Canny edge detection algorithm) are used to identify the edge positions of the water level markings. Image feature extraction methods (such as Hough transform or SIFT) are used to identify the specific positions of the water level markings (such as dam scale lines). The water level height is calculated based on the vertical position changes of the markings.

[0054] S12: Acquire UAV remote sensing data and perform image registration, water surface extraction, and water level inversion processing, further including: S121: Perform geometric registration of water surface areas in remote sensing images taken at different times and angles.

[0055] This invention ensures data consistency in the water surface area by geometrically registering remote sensing images taken at different times and angles.

[0056] S122: Extract water surface areas from remote sensing images, invert water level height using image analysis, and estimate water level changes by combining historical topographic data and reservoir capacity curves.

[0057] This invention utilizes remote sensing images to extract water surface areas, inverts water level height using image analysis methods (such as pixel classification), and estimates water level changes by combining existing topographic data and reservoir capacity curves.

[0058] S13: Acquire water level sensor monitoring data, perform time series smoothing and error correction, and filter outliers.

[0059] For data collected by traditional water level sensors, this embodiment of the invention performs time series smoothing and error correction on the acquired water level sensor monitoring data, filters out outliers, and ensures that the water level data output by the sensor is accurate.

[0060] Specifically, step S2 involves fusing the processed multi-source water level monitoring data and calculating the actual water level after data fusion, which further includes: S21: Construct a multi-source data quality assessment system and dynamically calculate the first confidence weight of each data source.

[0061] S22: Optimize the first confidence weight by combining the entropy method and expert experience weight to obtain the second confidence weight.

[0062] S23: Based on the second confidence weight, the processed multi-source water level monitoring data are fused to calculate the actual water level after the fusion of multiple data sources.

[0063] Furthermore, step S21, which involves constructing a multi-source data quality assessment system and dynamically calculating the first confidence weight of each data source, also includes: S211: Calculate the residual standard deviation of each data source over N time periods compared to the fusion result. The calculation formula is as follows: (1) In the formula, Let be the observation value of the i-th data source at time k. This is the result of the fusion.

[0064] S212: Calculate the deviation of each data source from the median value of the multi-source data at the current time. The calculation formula is as follows: (2) In the formula, Let be the observation value of the i-th data source at time t. Let t be the median value of the multi-source data at time t.

[0065] S213: Based on residual standard deviation and degree of deviation Calculate the first confidence weight for each data source. The calculation formula is as follows: (3) In the formula, and To adjust the parameters, M represents the data source.

[0066] Furthermore, step S22, which combines entropy method and expert experience weights to optimize the first confidence weight, to obtain the second confidence weight, also includes: A sliding window mechanism is used to dynamically update the weight allocation strategy. Combining the entropy method and expert experience weights, the first confidence weight is optimized through game theory to obtain the second confidence weight. The calculation formula is as follows: (4) In the formula, For objective weighting, Subjective weighting, This is the adjustment coefficient.

[0067] In a specific embodiment of the present invention, a sliding window mechanism is first used to dynamically update the weight allocation strategy, the sliding window length T is set, and the error characteristics of each data source within the window are stored.

[0068] Then, the objective weights are calculated based on the entropy method. The calculation process is as follows: (5) (6) (7) in, Let be the information entropy of the i-th data source, reflecting the stability of that data source within window T. The lower the entropy value, the smaller the data collision. T represents the sliding window length, used to calculate the number of historical time periods for weights, reflecting the time scale of weight updates. is the real-time deviation of the i-th data source at time k, used to measure the difference between this data source and the median value of multiple sources; Let be the deviation percentage of the i-th data source at time k, representing the proportion of the deviation of that data source to the total deviation. ; It is the natural logarithm of the sliding window time, used for standardizing entropy calculation, and is dimensionless; The objective weight of the i-th data source is calculated based on the entropy method and is entirely determined by the statistical characteristics of the data. ,and .

[0069] Finally, the weighting of expert experience is considered. The final weight, namely the second confidence weight, is obtained through game theory combinatorial optimization. Its calculation formula is as follows: (8) Solving for:

[0070] in, This is the final weight vector, used to balance objective statistics and subjective experience; The objective weight vector is derived from the entropy method. ; This is the subjective weight vector, which is the baseline weight set by experts based on prior knowledge; This is an adjustment coefficient used to control the balance between objective and subjective weights. It is determined through training with historical data that when Completely dependent on objective weights (data-driven), when It relies entirely on subjective weighting (expert experience-driven).

[0071] Furthermore, step S23 performs data fusion on the processed multi-source water level monitoring data based on the second confidence weight, and calculates the actual water level after fusion of multiple data sources. The calculation formula is as follows: (9) In the formula, z is the actual water level value after multi-source data fusion; These are the water level values ​​monitored by each data source; The second confidence weights for each data source.

[0072] This invention combines high-point video image recognition technology, UAV remote sensing water level monitoring data, and traditional water level sensor data to construct a two-layer optimization architecture that includes dynamic reliability assessment and adaptive weight allocation. This provides a real-time, accurate, and fault-tolerant water level monitoring method. Especially during flood season or in complex environments, traditional water level monitoring equipment is easily interfered with and cannot accurately reflect water level fluctuations. This invention, through video image processing technology and UAV remote sensing data analysis, overcomes the shortcomings of traditional equipment in situations with drastic water level fluctuations, equipment failures, or complex environments, significantly improving the accuracy and stability of water level monitoring. By establishing a multi-source data quality dynamic quantification adaptive fusion mechanism, an error backpropagation channel is created between the fusion result and the measured true value, thereby achieving dynamic adjustment and fusion of weight parameters and improving the accuracy, stability, and applicability of water level monitoring.

[0073] Specifically, step S3, which calculates the inflow rate based on the actual water level and real-time outflow data, includes: S31: The change in reservoir capacity is calculated based on the water level-capacity curve. The calculation formula is as follows: (10) (11) Where V is the reservoir capacity and z is the actual water level. The relationship between water level and reservoir capacity is shown in the curve. This represents the change in reservoir capacity during the scheduling period t.

[0074] S32: Obtain the real-time outflow from the flow meter, calculate and output the real-time inflow using the water balance formula. The calculation formula is as follows: (12) in, This refers to the inbound flow rate; t represents the outbound flow rate, and t represents the scheduling period.

[0075] Specifically, step S4 inputs the inflow into the hydrological forecasting model for dynamic assimilation, generating flood forecast results, which further include: S41: Initialize the hydrological forecasting model based on historical water level monitoring data to generate the model's forecasting boundary conditions. The historical water level monitoring data includes historical water level, flow rate, and rainfall data. The forecasting boundary conditions can be expressed as: (13) In the formula, The initial water level, This represents the initial outbound flow. This is the initial storage capacity.

[0076] S42: Real-time acquisition of water level monitoring data and outflow, and calculation of inflow based on the water balance formula.

[0077] This invention embodiment acquires the water level at time t in real time using a high-point video water level monitoring device. The outbound flow rate at time t is obtained in real time through flow monitoring equipment. Then, the inflow rate is calculated according to the water balance formula (12). The calculation formula is as follows:

[0078] (14) in, Changes in storage capacity can be represented by a storage capacity curve. Calculations of water level changes yielded the following results: for.

[0079] S43: Input the calculated inflow as assimilation information into the hydrological forecasting model to generate flood prediction values.

[0080] This embodiment of the invention assumes that the inflow predicted by the hydrological forecasting model is The observed inflow rate was The assimilated inbound flow is By adjusting the weight K t The corrected model yields: (15) (16) Among them, K t For Kalman gain, P t and R t These are the model error covariance and the observation error covariance, respectively.

[0081] Run the hydrological forecasting model based on the revised state to generate a new flood forecast value X. t+k : (17) Where k is the prediction step size, X t Let Q be the boundary condition of the model at time t. in,t+k Q represents the inbound flow at time t+k. out,t+k This represents the outbound flow rate at time t+k.

[0082] S44: Based on the error between the flood forecast and the actual values, the assimilation parameters in the model are updated in real time, where the error between the flood forecast and the actual values ​​is... The calculation formula is: (18) This invention calculates the prediction error and adjusts and dynamically updates the Kalman gain in real time. This improves the assimilation effect and ultimately enhances the accuracy and real-time performance of flood forecasts.

[0083] S45: Based on the updated assimilation parameters, dynamically revise the hydrological forecast model to generate optimal flood forecast results.

[0084] This invention uses inflow data derived from the water balance formula as assimilated data, inputting it into a hydrological forecasting model in real time. Combined with dynamic adjustment mechanisms such as Kalman filtering, the model's state is corrected, significantly reducing forecast errors and improving the accuracy and timeliness of flood forecasts. Compared to traditional hydrological forecasting models, which have weak assimilation capabilities for real-time flow and often rely on fixed boundary conditions or lagged data, failing to respond promptly to changes, this invention dynamically assimilates inflow information, enabling real-time adjustment of the model's state and improving forecast accuracy. Particularly during flood warning periods, it greatly enhances the reservoir's real-time flood control capabilities.

[0085] Example 2 Based on the above method, embodiments of the present invention also provide a real-time flood forecasting system for reservoirs based on multi-source data fusion technology, such as... Figures 8-11 As shown, it includes: The data acquisition and processing module 01 is used to acquire and preprocess multi-source water level monitoring data, which includes high-point video monitoring data, UAV remote sensing data and water level sensor monitoring data.

[0086] The data fusion and calculation module 02 is used to fuse the processed multi-source water level monitoring data and calculate the actual water level after data fusion.

[0087] The flow rate calculation module 03 is used to calculate the inflow rate based on the actual water level and real-time outflow data.

[0088] The data assimilation and prediction module 04 is used to input the inflow into the hydrological forecasting model for dynamic assimilation and generate the optimal flood forecast results.

[0089] Specifically, the data fusion and computing module 01 also includes: The first calculation module 011 is used to construct a multi-source data quality assessment system and dynamically calculate the first confidence weight of each data source.

[0090] The second calculation module 012 is used to optimize the first confidence weight by combining the entropy method and expert experience weight to obtain the second confidence weight.

[0091] The third calculation module 013 is used to perform data fusion on the processed multi-source water level monitoring data based on the second confidence weight, and calculate the actual water level after the fusion of multiple data sources.

[0092] In this embodiment of the invention, the first calculation module 011 first calculates the residual standard deviation of each data source and the fusion result over N time periods. Then, calculate the degree of deviation between each data source and the median of the multi-source data at the current time. Finally, based on the residual standard deviation and degree of deviation Calculate the first confidence weight for each data source. Then, the weight allocation strategy is dynamically updated through the second calculation module 012. Combining the entropy method and expert experience weights, the first confidence weight is optimized through game theory to obtain the second confidence weight. Finally, the third calculation module 013 performs data fusion on the processed multi-source water level monitoring data to calculate the actual water level after fusion of multiple data sources.

[0093] Specifically, the traffic back-tracking module 03 also includes: The reservoir capacity calculation module 031 is used to calculate the change in reservoir capacity based on the water level-capacity curve.

[0094] The inflow calculation module 032 is used to obtain the real-time outflow from the flow meter, calculate and output the real-time inflow using the water balance formula.

[0095] In this embodiment of the invention, the change in reservoir capacity is first calculated by the reservoir capacity calculation module 031, and then the real-time outflow is obtained from the flow meter by the inflow calculation module 032. The real-time inflow is calculated and output by combining the water balance formula.

[0096] Specifically, the data assimilation and prediction module 04 also includes: Model initialization module 041 is used to initialize the hydrological forecasting model based on historical water level monitoring data, which includes historical water level, flow rate and rainfall data.

[0097] The calculation module 042 is used to acquire water level monitoring data and outflow in real time, and calculate inflow according to the water balance formula.

[0098] The data assimilation module 043 is used to input the calculated inflow as assimilation information into the hydrological forecasting model to generate flood prediction values.

[0099] The parameter update module 044 is used to provide real-time feedback and dynamically update the assimilation parameters in the model based on the error between the flood prediction value and the actual value.

[0100] Prediction module 045 is used to dynamically revise the hydrological forecast model based on updated assimilation parameters to generate optimal flood forecast results.

[0101] This invention first initializes the hydrological forecasting model using the model initialization module 041, generating forecast boundary conditions. Then, the calculation module 042 obtains the water level and outflow at time t, calculates the inflow according to the water balance formula, assimilates the inflow using the data assimilation module 043, dynamically updates the assimilation parameters in the model using the parameter update module 044, and finally corrects the hydrological forecasting model using the prediction module 045 to generate the optimal flood forecast result. This invention uses the inflow calculated from the water balance formula as assimilation data, inputs it into the hydrological forecasting model in real time, and combines it with dynamic adjustment mechanisms such as Kalman filtering to correct the state of the hydrological model, significantly reducing forecast errors and improving the accuracy and timeliness of flood forecasts. Compared to traditional hydrological forecasting models, which have weak assimilation capabilities for real-time flow and often rely on fixed boundary conditions or lagging data, making them unable to respond promptly to changes, this invention dynamically assimilates inflow information, enabling real-time adjustment of the model state and improving forecast accuracy. Especially during flood warning periods, it greatly enhances the real-time flood control capabilities of reservoirs.

[0102] Example 3 like Figure 12 As shown, this embodiment of the invention also provides an electronic device, including a processor and a memory. The memory stores a computer program, which is loaded and executed by the processor to implement the above-described real-time flood forecasting method for reservoirs based on multi-source data fusion technology. The device in this invention can be a server, PC, PAD, mobile phone, etc.

[0103] Furthermore, this embodiment of the invention also provides a computer-readable storage medium storing a computer program, which is loaded and executed by a processor to implement the above-described real-time flood forecasting method for reservoirs based on multi-source data fusion technology.

[0104] In summary, the real-time flood forecasting method and system for reservoirs based on multi-source data fusion technology described in this invention has the following advantages: 1) Improve the reliability and accuracy of water level monitoring data. This invention integrates high-point video image recognition technology, UAV remote sensing data, and traditional water level sensor data to provide high-precision water level monitoring data in complex reservoir environments and under special conditions such as flood season. By calculating objective weights using the entropy method and combining expert experience, a game theory-based combinatorial optimization formula is constructed, thereby achieving effective fusion of multi-source water level monitoring data. This effectively avoids errors that may result from a single data source and can reflect dynamic changes in water levels in real time and accurately, thus improving the reliability and accuracy of water level monitoring and further enhancing the accuracy of reservoir flood forecasting.

[0105] 2) Accurately calculate inbound flow rate to improve flow monitoring accuracy. This invention achieves accurate calculation of inflow by precisely acquiring water level and outflow data and combining them with the water balance formula to inversely deduce inflow. This technology can provide inflow data in real time and accurately, serving as an important boundary condition for hydrological forecasting models. It effectively reduces forecast deviations caused by inflow errors, improving the accuracy of flow monitoring and the reliability of reservoir scheduling.

[0106] 3) Real-time assimilation of inbound traffic reduces model error. This invention uses real-time inflow data as assimilated into a hydrological forecasting model, enabling dynamic adjustment of the model's state and reducing errors. Specifically, the Kalman filter algorithm dynamically optimizes model weights, allowing the hydrological forecasting model to make accurate corrections based on real-time data in rapidly changing flood scenarios. This technology effectively improves the accuracy and real-time performance of hydrological models, further enhancing the reliability of flood forecasts and strengthening the real-time flood control capabilities and response speed of reservoirs.

[0107] 4) Improve the timeliness and accuracy of flood warnings. This invention provides more comprehensive and accurate input data for hydrological forecasting models through multi-source data fusion, enabling the models to more accurately capture key changes during flood processes and improving the timeliness and accuracy of early warnings. Combined with assimilation technology, the system can adjust and optimize forecast results in real time, reducing the risks associated with false alarms or missed alarms. As a result, reservoir management departments can obtain more reliable flood warning information in a timely manner, thereby effectively formulating flood control decisions and minimizing flood losses.

[0108] Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application.

Claims

1. A method for real-time flood forecasting of reservoirs based on multi-source data fusion technology, characterized in that, include: Acquire multi-source water level monitoring data and perform preprocessing; The processed multi-source water level monitoring data are fused, and the actual water level after data fusion is calculated. The inflow rate is calculated based on the actual water level and real-time outflow data. The inflow is input into the hydrological forecasting model for dynamic assimilation to generate flood forecast results.

2. The method for real-time flood forecasting of reservoirs based on multi-source data fusion technology according to claim 1, characterized in that, The multi-source water level monitoring data includes high-point video monitoring data, UAV remote sensing data, and water level sensor monitoring data.

3. The method for real-time flood forecasting of reservoirs based on multi-source data fusion technology according to claim 2, characterized in that, The acquisition and preprocessing of multi-source water level monitoring data further includes: Acquire high-point video monitoring data and perform image denoising, edge detection, and marker localization; Acquire UAV remote sensing data and perform image registration, water surface extraction, and water level inversion processing; The system acquires water level sensor monitoring data, performs time series smoothing and error correction, and filters outliers.

4. The method for real-time flood forecasting of reservoirs based on multi-source data fusion technology according to claim 3, characterized in that, The process of acquiring UAV remote sensing data and performing image registration, water surface extraction, and water level inversion further includes: Geometric registration of water surface areas in remote sensing images taken at different times and angles; Water surface areas are extracted from remote sensing images, water level height is retrieved through image analysis, and water level changes are estimated by combining historical topographic data and reservoir capacity curves.

5. The method for real-time flood forecasting of reservoirs based on multi-source data fusion technology according to claim 1, characterized in that, The step of fusing the processed multi-source water level monitoring data and calculating the actual water level after data fusion further includes: Construct a multi-source data quality assessment system and dynamically calculate the first confidence weight of each data source; The first confidence weight is optimized by combining the entropy method and expert experience weight to obtain the second confidence weight; The processed multi-source water level monitoring data are fused based on the second confidence weight to calculate the actual water level after the fusion of multiple data sources.

6. The method for real-time flood forecasting of reservoirs based on multi-source data fusion technology according to claim 5, characterized in that, The construction of a multi-source data quality assessment system, and the dynamic calculation of the first confidence weight of each data source, further include: Calculate the residual standard deviation of each data source and the fusion result over N time periods. The calculation formula is as follows: In the formula, Let be the observation value of the i-th data source at time k. The result of the fusion; Calculate the degree of deviation between each data source and the median of the multi-source data at the current time. The calculation formula is as follows: In the formula, Let be the observation value of the i-th data source at time t. The median value of the multi-source data at time t; Based on residual standard deviation and degree of deviation Calculate the first confidence weight for each data source. The calculation formula is as follows: In the formula, and To adjust the parameters, M represents the data source.

7. The method for real-time flood forecasting of reservoirs based on multi-source data fusion technology according to claim 5, characterized in that, The optimization of the first confidence weight by combining the entropy method and expert experience weights to obtain the second confidence weight further includes: A sliding window mechanism is used to dynamically update the weight allocation strategy. Combining the entropy method and expert experience weights, the first confidence weight is optimized through game theory to obtain the second confidence weight. The calculation formula is as follows: In the formula, For objective weighting, Subjective weighting, This is the adjustment coefficient.

8. The method for real-time flood forecasting of reservoirs based on multi-source data fusion technology according to claim 7, characterized in that, The formula for calculating the actual water level is: In the formula, z is the actual water level value after multi-source data fusion; These are the water level values ​​monitored by each data source; The second confidence weights for each data source.

9. The method for real-time flood forecasting of reservoirs based on multi-source data fusion technology according to claim 1, characterized in that, The step of deriving the inflow rate based on the actual water level and real-time outflow data further includes: The change in reservoir capacity is calculated based on the water level-capacity curve, and the formula is as follows: Where V is the reservoir capacity and z is the actual water level. The relationship between water level and reservoir capacity is shown in the curve. This represents the change in reservoir capacity within the scheduling period t. The real-time outflow rate is obtained from the flow meter, and the real-time inflow rate is calculated and output using the water balance formula. The calculation formula is as follows: in, This refers to the inbound flow rate; t represents the outbound flow rate, and t represents the scheduling period.

10. The method for real-time flood forecasting of reservoirs based on multi-source data fusion technology according to claim 1, characterized in that, The step of inputting the inflow into the hydrological forecasting model for dynamic assimilation to generate flood forecasting results further includes: The hydrological forecasting model is initialized based on historical water level monitoring data, which includes historical water level, flow rate and rainfall data. Real-time acquisition of water level monitoring data and outflow rate, and calculation of inflow rate based on water balance formula; The calculated inflow is used as assimilated information and input into the hydrological forecasting model to generate flood prediction values; Based on the error between the flood forecast and the actual value, the assimilation parameters in the model are fed back in real time and dynamically updated. The hydrological forecasting model is dynamically revised based on the updated assimilation parameters to generate optimal flood forecasting results.

11. A real-time flood forecasting system for reservoirs based on multi-source data fusion technology, characterized in that, include: The data acquisition and processing module is used to acquire and preprocess multi-source water level monitoring data; The data fusion and calculation module is used to fuse the processed multi-source water level monitoring data and calculate the actual water level after data fusion. The flow rate calculation module is used to calculate the inflow rate based on the actual water level and real-time outflow data. The data assimilation and prediction module is used to dynamically assimilate the inflow into the hydrological forecasting model and generate the optimal flood forecast results.

12. The method for real-time flood forecasting of reservoirs based on multi-source data fusion technology according to claim 11, characterized in that, The data fusion and computation module also includes: The first calculation module is used to construct a multi-source data quality assessment system and dynamically calculate the first confidence weight of each data source. The second calculation module is used to optimize the first confidence weight by combining the entropy method and expert experience weight to obtain the second confidence weight; The third calculation module is used to perform data fusion on the processed multi-source water level monitoring data based on the second confidence weight, and calculate the actual water level after the fusion of multiple data sources.

13. The method for real-time flood forecasting of reservoirs based on multi-source data fusion technology according to claim 11, characterized in that, The flow inversion module also includes: The reservoir capacity calculation module is used to calculate the change in reservoir capacity based on the water level-capacity curve. The inflow calculation module is used to obtain the real-time outflow from the flow meter, calculate and output the real-time inflow using the water balance formula.

14. The method for real-time flood forecasting of reservoirs based on multi-source data fusion technology according to claim 11, characterized in that, The data assimilation and prediction module also includes: The model initialization module is used to initialize the hydrological forecasting model based on historical water level monitoring data, wherein the historical water level monitoring data includes historical water level, flow rate and rainfall data; The calculation module is used to acquire water level monitoring data and outflow in real time, and calculate inflow according to the water balance formula. The data assimilation module is used to input the calculated inflow as assimilation information into the hydrological forecasting model to generate flood prediction values; The parameter update module is used to provide real-time feedback and dynamically update the assimilation parameters in the model based on the error between the flood forecast and the actual values. The prediction module is used to dynamically revise the hydrological forecasting model based on updated assimilation parameters to generate optimal flood forecast results.

15. An electronic device, characterized in that, include: A processor and a memory, wherein the memory stores a computer program, which is loaded and executed by the processor to implement the real-time flood forecasting method for reservoirs based on multi-source data fusion technology as described in any one of claims 1 to 10.

16. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program, which is loaded and executed by a processor to implement the real-time flood forecasting method for reservoirs based on multi-source data fusion technology as described in any one of claims 1 to 10.

Images