Intelligent water conservancy monitoring method and system based on multi-source data fusion

Abstract

The application provides a smart water conservancy monitoring method and system based on multi-source data fusion, and relates to the field of data processing.The method comprises the following steps: determining a plurality of inflow monitoring points of a reservoir according to historical inflow data of the reservoir; setting multi-source data acquisition devices at each inflow monitoring point; determining an optimal edge computing architecture and an optimal network architecture according to the plurality of inflow monitoring points; for each inflow monitoring point, determining the heartbeat data packet upload frequency of the inflow monitoring point according to the multi-source data collected by the multi-source data acquisition devices of the inflow monitoring point through the optimal edge computing architecture; uploading the multi-source data to a water conservancy monitoring platform according to the heartbeat data packet upload frequency of each inflow monitoring point through the optimal network architecture; and the water conservancy monitoring platform fuses the multi-source data collected by the multi-source data acquisition devices of the plurality of inflow monitoring points to perform reservoir water level prediction, thereby improving the timeliness and intelligent level of water conservancy monitoring.

Description

Technical Field

[0001] The present invention relates to the field of data processing, and in particular to a smart water conservancy monitoring method and system based on multi-source data fusion. Background Art

[0002] Water conservancy monitoring refers to a system that uses modern information technology (such as sensors, communications, the Internet of Things, big data, and artificial intelligence) to conduct real-time monitoring, data collection, transmission, analysis, and decision support for water bodies, water conservancy facilities, and the water environment. Its core goals are to achieve scientific management of water resources, effective prevention and control of water disasters, and sustainable protection of aquatic ecosystems.

[0003] Existing water monitoring technologies often rely on non-real-time data processing and analysis, lacking the ability to instantly reflect the status of water systems. Static data analysis methods often struggle to handle rapidly changing hydrological events, failing to provide immediate data updates and decision support. Furthermore, existing technologies experience a lag in integrating new data, limiting the timeliness and accuracy of water resource management decisions and increasing uncertainty and risk in the management process.

[0004] Therefore, it is necessary to provide a smart water conservancy monitoring method and system based on multi-source data fusion to improve the timeliness and intelligence level of water conservancy monitoring. Summary of the Invention

[0005] The present invention provides a smart water conservancy monitoring method based on multi-source data fusion, including: determining multiple water inflow monitoring points of the reservoir based on the historical water inflow data of the reservoir; setting a multi-source data acquisition device at each water inflow monitoring point; determining the optimal edge computing architecture and the optimal network architecture based on the multiple water inflow monitoring points; for each water inflow monitoring point, determining the heartbeat data packet upload frequency of the water inflow monitoring point through the optimal edge computing architecture based on the multi-source data collected by the multi-source data acquisition device of the water inflow monitoring point; uploading the multi-source data collected by the multi-source data acquisition device of the water inflow monitoring point to a water conservancy monitoring platform through the optimal network architecture based on the heartbeat data packet upload frequency of each water inflow monitoring point; the water conservancy monitoring platform fuses the multi-source data collected by the multi-source data acquisition devices of the multiple water inflow monitoring points to predict the reservoir water level.

[0006] Furthermore, based on the historical water inflow data of the reservoir, multiple water inflow monitoring points of the reservoir are determined, including: obtaining a regional digital elevation model of the reservoir; determining multiple initial monitoring points based on the regional digital elevation model of the reservoir; obtaining historical water inflow data of the multiple initial monitoring points and historical water level data of the reservoir; determining the correlation coefficient between the water inflow status of the initial monitoring points and the water level of the reservoir based on the historical water inflow data of the multiple initial monitoring points and the historical water level data of the reservoir; and determining multiple water inflow monitoring points of the reservoir based on the correlation coefficient between the water inflow status of each initial monitoring point and the water level of the reservoir through a genetic algorithm.

[0007] Furthermore, a genetic algorithm is used to determine multiple water inflow monitoring points of the reservoir based on the correlation coefficient between the water inflow status of each initial monitoring point and the water level of the reservoir, including: calculating the water inflow status similarity and the water inflow status correlation coefficient of two initial monitoring points based on the historical water inflow data of the multiple initial monitoring points; sampling from the multiple initial monitoring points to initialize multiple individuals, wherein each individual includes at least three initial monitoring points; establishing a fitness function, wherein the fitness function is related to the water inflow status similarity and the water inflow status correlation coefficient of any two initial monitoring points included in the individual and the correlation coefficient between the water inflow status of any initial monitoring point and the water level of the reservoir; calculating the fitness value of each individual based on the fitness function; and repeating selection, crossover, mutation and termination condition judgment on multiple individuals based on the fitness value of each individual until the multiple water inflow monitoring points of the reservoir are determined.

[0008] Furthermore, based on multiple water inflow monitoring points, the optimal edge computing architecture and the optimal network architecture are determined, including: determining the water inflow status correlation coefficient of any two water inflow monitoring points based on the historical water inflow data of any two water inflow monitoring points; clustering the multiple water inflow monitoring points according to the water inflow status correlation coefficient, spatial distance, correlation coefficient between the water inflow status of each water inflow monitoring point and the water level of the reservoir, and the constraint on the number of water inflow monitoring points to determine multiple clusters; based on the multiple clusters, the optimal edge computing architecture and the optimal network architecture are determined.

[0009] Furthermore, a plurality of water inflow monitoring points are clustered according to the water inflow status correlation coefficient of any two water inflow monitoring points, the spatial distance, the correlation coefficient between the water inflow status of each water inflow monitoring point and the water level of the reservoir, and the water inflow monitoring point quantity constraint, to determine a plurality of clusters, including: S11, screening a plurality of water inflow monitoring points as cluster centers according to the water inflow status correlation coefficient of any two water inflow monitoring points; S12, sorting the water inflow monitoring points that are not cluster centers according to the correlation coefficient between the water inflow status of the water inflow monitoring points and the water level of the reservoir, and generating a sorting result; S13, determining the water inflow monitoring points of the current cluster according to the sorting result; S14, determining the water inflow monitoring points of the current cluster according to the number of water inflow monitoring points included in each cluster and the water inflow monitoring point quantity constraint, and the water inflow monitoring points of the current cluster and each cluster The spatial distance between the centers is used to determine the candidate cluster centers, and the candidate cluster center with the smallest water status correlation coefficient is used as the cluster center of the water inlet monitoring point of the current cluster for clustering; S15, update the number of water inlet monitoring points included in each cluster; S16, determine whether all water inlet monitoring points that are not cluster centers have completed clustering, if so, execute S18, if not, execute S17; S17, according to the sorting result, use the next water inlet monitoring point as the water inlet monitoring point of the current cluster; S18, determine whether the iteration end condition is met, if so, complete clustering and output multiple clusters, if not, execute S19; S19, according to the number of water inlet monitoring points included in each cluster and the water status correlation coefficient of any two water inlet monitoring points included in each cluster, update the cluster center and execute S12.

[0010] Furthermore, the multi-source data collected by the multi-source data acquisition device of the water inlet monitoring point includes at least water level sequence, flow velocity sequence and image data; through the optimal edge computing architecture, the heartbeat data packet upload frequency of the water inlet monitoring point is determined according to the multi-source data collected by the multi-source data acquisition device of the water inlet monitoring point, including: calculating the water level anomaly value of the water inlet monitoring point according to the water level sequence; calculating the flow velocity anomaly value of the water inlet monitoring point according to the flow velocity sequence; calculating the color anomaly value of the water inlet monitoring point according to the image data; calculating the inflow water anomaly value of the water inlet monitoring point according to the water level anomaly value, flow velocity anomaly value and color anomaly value of the water inlet monitoring point; determining the heartbeat data packet upload frequency of the water inlet monitoring point according to the inflow water anomaly value of the water inlet monitoring point.

[0011] Furthermore, the water conservancy monitoring platform integrates multi-source data collected by multi-source data acquisition devices at multiple water inflow monitoring points to predict the reservoir water level, including: for each water inflow monitoring point, based on the water inflow status correlation coefficient of any two water inflow monitoring points, determining the relevant water inflow monitoring points of the water inflow monitoring point, and according to the relevant water inflow monitoring points of the water inflow monitoring point; based on the relevant water inflow monitoring points of each water inflow monitoring point and the multi-source data collected by the multi-source data acquisition devices at multiple water inflow monitoring points, correcting the water level sequence and flow rate sequence of the water inflow monitoring point; and predicting the reservoir water level based on the water level sequence and flow rate sequence of the corrected water inflow monitoring point.

[0012] Furthermore, based on the multi-source data collected by the multi-source data acquisition device of the relevant water inlet monitoring points and multiple water inlet monitoring points of each water inlet monitoring point, the water level sequence and the flow velocity sequence of the water inlet monitoring point are corrected, including: S21, performing variational mode decomposition on the water level sequence and the flow velocity sequence of the water inlet monitoring point to obtain the intrinsic mode function and residual of the water level sequence and the intrinsic mode function and residual of the flow velocity sequence; S22, for each water inlet monitoring point, the intrinsic mode function and residual of the water level sequence of the relevant water inlet monitoring point and the intrinsic mode function and residual of the flow velocity sequence, the intrinsic mode function and residual of the water level sequence and the flow velocity sequence of the water inlet monitoring point are decomposed. The intrinsic mode functions and residuals of the sequence are corrected; S23. Based on the intrinsic mode functions and residuals of the water level sequence before correction and the intrinsic mode functions and residuals of the flow velocity sequence and the intrinsic mode functions and residuals of the water level sequence after correction and the intrinsic mode functions and residuals of the flow velocity sequence of each inflow monitoring point, the global correction change value is calculated; S24. It is determined whether the global correction change value converges. If so, the correction is completed. If not, S25 is executed; S25. For each inflow monitoring point, based on the intrinsic mode functions and residuals of the water level sequence after correction and the intrinsic mode functions and residuals of the flow velocity sequence, data is updated and S22 is executed.

[0013] Furthermore, the water conservancy monitoring platform is also used to determine the geological disaster warning area based on the color anomaly values ​​of multiple water inflow monitoring points.

[0014] The present invention provides a smart water conservancy monitoring system based on multi-source data fusion, including: a data acquisition module, which is used to determine multiple water inflow monitoring points of a reservoir based on the historical water inflow data of the reservoir, and set a multi-source data acquisition device at each water inflow monitoring point; a data transmission module, which is used to determine the optimal edge computing architecture and the optimal network architecture based on the multiple water inflow monitoring points, and for each water inflow monitoring point, through the optimal edge computing architecture, according to the multi-source data collected by the multi-source data acquisition device of the water inflow monitoring point, determine the heartbeat data packet upload frequency of the water inflow monitoring point, and through the optimal network architecture, according to the heartbeat data packet upload frequency of each water inflow monitoring point, upload the multi-source data collected by the multi-source data acquisition device of the water inflow monitoring point to a water conservancy monitoring platform; a water level monitoring module, which is used to fuse the multi-source data collected by the multi-source data acquisition devices of multiple water inflow monitoring points through the water conservancy monitoring platform to predict the reservoir water level.

[0015] Compared with the existing technology, the intelligent water conservancy monitoring method and system based on multi-source data fusion provided by the present invention have at least the following beneficial effects:

[0016] 1. Multiple water inflow monitoring points are determined based on historical reservoir inflow data. The layout of these monitoring points is more scientific and targeted, covering key areas of reservoir inflow, avoiding monitoring blind spots, and ensuring accurate monitoring of reservoir inflow conditions. Multi-source data acquisition devices can flexibly adjust monitoring parameters and frequency based on changes in reservoir inflow. For example, during periods of heavy rainfall, the frequency of water level and flow monitoring can be increased to provide timely insight into dynamic changes in reservoir inflow. The water conservancy monitoring platform integrates multi-source data collected by multi-source data acquisition devices at multiple water inflow monitoring points to comprehensively consider the impact of various factors on reservoir water levels. For example, data such as rainfall, flow, and flow velocity have complex correlations with water levels. Multi-source data fusion enables a more accurate water level prediction model to be developed, improving the accuracy of water level predictions. Accurate water level predictions provide a scientific basis for reservoir operation and management. Based on prediction results, water conservancy departments can formulate emergency plans for flood control, drought relief, and other emergencies in advance, rationally arrange reservoir water storage and release, reduce operational risks, and ensure the safety and stability of the surrounding areas.

[0017] 2. Obtaining a regional digital elevation model of the reservoir provides a visual representation of the surrounding topography. Multiple initial monitoring points, determined based on the regional digital elevation model, fully account for the impact of topography on water inflow. For example, setting initial monitoring points in valleys and gullies, where water flow tends to converge, allows for more accurate capture of water inflow conditions in different areas, avoids monitoring blind spots caused by topography, and ensures a more scientific and rational layout of monitoring points. By calculating the correlation coefficient between the water inflow status of initial monitoring points and the reservoir water level, the degree of influence of the water inflow at each monitoring point on the reservoir water level can be determined. The higher the correlation coefficient, the closer the relationship between water inflow changes at that monitoring point and reservoir water level changes, and the greater its reference value for reservoir water level prediction and management. Using a genetic algorithm, multiple individuals (composed of initial monitoring points) are selected, crossover, and mutated based on the similarity of water inflow status, the correlation coefficient, and the correlation coefficient with the reservoir water level. This allows for the identification of the optimal combination of water inflow monitoring points. This method can avoid the subjectivity and limitations of manually determining monitoring points, ensuring that the layout of monitoring points can not only fully reflect the water inflow situation of the reservoir, but also efficiently utilize monitoring resources.

[0018] 3. Clustering is performed based on the water inflow status correlation coefficient, spatial distance, correlation coefficient between the water inflow status of each water inflow monitoring point and the reservoir water level, and the quantity constraint of water inflow monitoring points. The monitoring points can be grouped comprehensively and scientifically. Specifically, the water inflow monitoring points with a higher water inflow status correlation coefficient have similar fluctuation patterns and change laws in the time series. When the upload frequency of the heartbeat data packet of a certain water inflow monitoring point increases, the upload frequency of the water inflow monitoring point with a higher water inflow status correlation coefficient is also likely to increase, resulting in a surge in the computing power demand of the same edge computing node. The water inflow monitoring points with a smaller water inflow status correlation coefficient are divided into a cluster, which effectively avoids the surge in computing power demand of the same edge computing node, resulting in congestion in data processing and transmission. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] This specification will be further described in the form of exemplary embodiments, which will be described in detail with reference to the accompanying drawings. These embodiments are not limiting, and in these embodiments, like numbers represent like structures, wherein:

[0020] Figure 1 This is a flowchart of a smart water conservancy monitoring method based on multi-source data fusion according to some embodiments of this specification;

[0021] Figure 2 is a schematic diagram of a process for determining multiple clusters according to some embodiments of this specification;

[0022] Figure 3This is a schematic diagram of a process for correcting the water level sequence and flow rate sequence of an incoming water monitoring point according to some embodiments of this specification;

[0023] Figure 4 It is a module diagram of a smart water conservancy monitoring system based on multi-source data fusion as shown in some embodiments of this specification. DETAILED DESCRIPTION

[0024] To more clearly illustrate the technical solutions of the embodiments of this specification, the following briefly describes the drawings required for describing the embodiments. Obviously, the drawings described below are merely examples or embodiments of this specification. Those skilled in the art can apply this specification to other similar scenarios based on these drawings without inventive effort. Unless otherwise apparent from the context or otherwise noted, the same reference numerals in the figures represent the same structure or operation.

[0025] Figure 1 This is a flow chart of a smart water conservancy monitoring method based on multi-source data fusion according to some embodiments of this specification, such as Figure 1 As shown, the smart water conservancy monitoring method based on multi-source data fusion can include the following steps.

[0026] Step 110: Determine multiple water inflow monitoring points of the reservoir based on historical water inflow data of the reservoir.

[0027] Specifically include:

[0028] Obtain regional digital elevation models of reservoirs;

[0029] Determine multiple initial monitoring points based on the regional digital elevation model of the reservoir;

[0030] Obtain historical water inflow data and historical water level data of multiple initial monitoring points;

[0031] Determine the correlation coefficient between the water inflow status of the initial monitoring point and the water level of the reservoir based on the historical water inflow data of multiple initial monitoring points and the historical water level data of the reservoir;

[0032] Through genetic algorithm, multiple water inflow monitoring points of the reservoir are determined according to the correlation coefficient between the water inflow status of each initial monitoring point and the water level of the reservoir.

[0033] Specifically, a Digital Elevation Model (DEM) is a model that represents the spatial distribution of actual terrain features in digital form, organized in a specific structure. It accurately describes ground elevation information through an ordered array of numerical values. DEMs for reservoir areas can be obtained through various means, including satellite remote sensing imagery, aerial photogrammetry, and digitized topographic maps. Using DEMs, terrain features such as elevation changes, slope, and aspect are analyzed around the reservoir. Using the regional DEM of the reservoir, hydrological algorithms (such as the D8 algorithm and the multi-flow direction algorithm) are used to calculate the flow direction for each grid cell, determine the surface flow path, and identify major river channels, tributaries, and potential confluence areas within the watershed. Initial monitoring points represent the topographic and hydrological characteristics of the area and reflect the primary contribution to the reservoir's water inflow. For example, initial monitoring points can be set at major river channels, tributary confluences, and low-lying areas. Multiple initial monitoring points can be determined empirically based on the regional DEM of the reservoir.

[0034] The historical water inflow data of the reservoir may include the historical water inflow data of multiple initial monitoring points and the historical water level data of the reservoir, wherein the historical water inflow data may include the water level and flow rate of the initial monitoring point at multiple historical time points, and the historical water level data of the reservoir may include the reservoir water level at the multiple historical time points.

[0035] For each initial monitoring point, the correlation coefficient between the water level of the initial monitoring point and the water level of the reservoir can be calculated by using the calculation formula of the correlation coefficient (for example, the Pearson correlation coefficient, the Spearman rank correlation coefficient, etc.) based on the water level of the initial monitoring point at multiple historical time points and the reservoir water level at these multiple historical time points. In the same way, the correlation coefficient between the flow velocity of the initial monitoring point and the water level of the reservoir is calculated. The correlation coefficient between the water level of the initial monitoring point and the water level of the reservoir and the correlation coefficient between the flow velocity and the water level of the reservoir are weightedly summed to calculate the correlation coefficient between the inflow status of the initial monitoring point and the water level of the reservoir.

[0036] Preferably, a genetic algorithm is used to determine multiple water inflow monitoring points of the reservoir based on the correlation coefficient between the water inflow status of each initial monitoring point and the water level of the reservoir, including:

[0037] Based on the historical water inflow data of multiple initial monitoring points, the similarity of the water inflow status and the water inflow status correlation coefficient of the two initial monitoring points are calculated;

[0038] Sampling is performed from a plurality of initial monitoring points to initialize a plurality of individuals, wherein each individual includes at least three initial monitoring points;

[0039] Establishing a fitness function, wherein the fitness function is related to the similarity of the water inflow state and the water inflow state correlation coefficient of any two initial monitoring points included in the individual, and the correlation coefficient between the water inflow state of any initial monitoring point and the water level of the reservoir;

[0040] Calculate the fitness value of each individual according to the fitness function;

[0041] According to the fitness value of each individual, multiple individuals are repeatedly selected, crossed, mutated and terminated until multiple water inflow monitoring points of the reservoir are determined.

[0042] Specifically, for any two initial monitoring points, the cosine distance of the water levels of the two initial monitoring points at multiple historical time points can be calculated, and the cosine distance of the flow velocities of the two initial monitoring points at multiple historical time points can be calculated. Based on the cosine distance of the water levels and the cosine distance of the flow velocities of the two initial monitoring points at multiple historical time points, the similarity of the inflow water states of the two initial monitoring points is calculated. Among them, the smaller the cosine distance of the water levels and the cosine distance of the flow velocities of the two initial monitoring points at multiple historical time points, the greater the similarity of the inflow water states of the two initial monitoring points.

[0043] The greater the mean of the similarity of the inflow status of any two initial monitoring points included in the individual, the smaller the value of the fitness function (i.e., fitness value); the greater the mean of the correlation coefficient between the inflow status of any initial monitoring point and the water level of the reservoir, the greater the value of the fitness function (i.e., fitness value).

[0044] As an example, the fitness function It can be:

[0045] ;

[0046] in, is the similarity of the water supply status between the i-th initial monitoring point and the j-th initial monitoring point included in the individual, is the number of initial monitoring points included in the individual, is the correlation coefficient between the initial monitoring point i and the water level of the reservoir, and is the weight, and greater than 0, ,For example, , .

[0047] Selection involves selecting individuals with higher fitness from the current population, giving them the opportunity to advance to the next generation, thereby preserving superior genes and improving the overall quality of the population. The probability of selection is calculated based on each individual's fitness value. Individuals with higher fitness values ​​have a greater probability of selection. The probability of selection can be calculated by dividing each individual's fitness value by the sum of the fitness values ​​of all individuals in the population. Individuals are then randomly selected by spinning a roulette wheel, with the probability of selection proportional to the area occupied by the wheel.

[0048] Crossover refers to the exchange of some genes between two individuals to generate new individuals, increase the diversity of the population, and help explore more solution spaces. For example, a parent individual contains the initial monitoring point A 、 B 、 C , another parent individual contains the monitoring point D 、 E 、 F , after the cross operation, it is possible to generate the initial monitoring points A 、 E 、 C offspring individuals.

[0049] Mutation refers to the random change of individual genes, the introduction of new gene combinations, to prevent the algorithm from falling into the local optimal solution and increase the diversity of the population. Since the individual genes are a combination of initial monitoring points, the replacement mutation method can be used, that is, randomly selecting an initial monitoring point in the individual and replacing it with other unselected initial monitoring points. For example, an individual contains monitoring points A 、 B 、 C , after the mutation operation, the monitoring point may be B Replace with monitoring point D , generate monitoring points A 、 D 、 C new individual.

[0050] Termination conditions may include reaching the maximum number of iterations (for example, 100 times), the fitness value reaching a preset threshold (for example, 0.8), the fitness value of the population no longer significantly improving (that is, in several consecutive iterations, the average fitness value or the best fitness value of the population no longer significantly improves), etc. For example, the algorithm is terminated when the average fitness value of the population changes by less than 0.01 in 10 consecutive iterations.

[0051] Output the initial monitoring points including the individuals with the highest fitness value in the current population as the final multiple water inflow monitoring points.

[0052] Step 120: Set up a multi-source data acquisition device at each water inlet monitoring point.

[0053] Specifically, the multi-source data collected by the multi-source data acquisition device at the water inflow monitoring point includes at least water level sequences, flow velocity sequences, and image data. Water level sensors (e.g., float-type water level gauges, pressure-type water level gauges, ultrasonic water level gauges, etc.) can be used to collect water levels, where the water level sequence can include water levels at multiple time points. Flow velocity sensors (e.g., propeller-type current meters, acoustic Doppler current profilers, etc.) can be used to collect flow velocity, where the flow velocity sequence can include flow velocity at multiple time points. High-definition cameras can be used to capture color image data of the water surface.

[0054] Step 130: Determine the optimal edge computing architecture and the optimal network architecture based on multiple water inflow monitoring points.

[0055] Specifically include:

[0056] Determine the water status correlation coefficient of any two water inflow monitoring points based on the historical water inflow data of any two water inflow monitoring points;

[0057] Clustering multiple inflow water monitoring points based on the water inflow status correlation coefficient and spatial distance between any two inflow water monitoring points, the correlation coefficient between the water inflow status of each inflow water monitoring point and the water level of the reservoir, and the constraint on the number of inflow water monitoring points (e.g., the maximum number is 10), to determine multiple clusters;

[0058] Based on multiple clusters, determine the optimal edge computing architecture and optimal network architecture.

[0059] Specifically, for any two water inflow monitoring points, the water level correlation coefficient of the two water inflow monitoring points can be calculated according to the water levels of the two water inflow monitoring points at multiple historical time points through the calculation formula of the correlation coefficient (for example, the Pearson correlation coefficient, the Spearman rank correlation coefficient, etc.). The flow velocity correlation coefficient of the two water inflow monitoring points can be calculated in the same way. The flow velocity correlation coefficient and the water level correlation coefficient of the two water inflow monitoring points are weightedly summed to obtain the water inflow status correlation coefficient of the two water inflow monitoring points.

[0060] The constraint on the number of water inflow monitoring points can be determined based on the computing power resources of the edge computing nodes and the computing power resources required to process the multi-source data of each water inflow monitoring point. For example, through actual testing or empirical estimation, the computing power resources such as the number of CPU cycles, memory bandwidth, and storage space required to process the multi-source data of each water inflow monitoring point can be determined, and the hardware resources of the edge computing nodes can be tested to obtain their available CPU cycles, memory bandwidth, and storage space. The maximum number of water inflow monitoring points can be calculated based on the data processing computing power requirements of a single monitoring point and the available computing power resources of the edge computing nodes.

[0061] Figure 2 is a schematic diagram of a process for determining multiple clusters according to some embodiments of this specification, such as Figure 2 As shown, as an example, multiple water inflow monitoring points are clustered according to the water inflow status correlation coefficient, spatial distance, correlation coefficient between the water inflow status of each water inflow monitoring point and the water level of the reservoir, and the number constraint of water inflow monitoring points to determine multiple clusters, including:

[0062] S11. Based on the water supply status correlation coefficient between any two water supply monitoring points, multiple water supply monitoring points are selected as cluster centers. For example, for each water supply monitoring point, the variance of the water supply status correlation coefficient between the water supply monitoring point and any other water supply monitoring point can be calculated, and the water supply monitoring point with a variance greater than a variance threshold is selected as the cluster center.

[0063] S12. Sort the water inflow monitoring points that are not the cluster center according to the correlation coefficient between the water inflow status of the water inflow monitoring points that are not the cluster center and the water level of the reservoir, and generate a sorting result. For example, sort the water inflow monitoring points that are not the cluster center according to the correlation coefficient between the water inflow status of the water inflow monitoring points that are not the cluster center and the water level of the reservoir from large to small.

[0064] S13. Determine the inlet water monitoring point of the current cluster according to the sorting result. For example, use the inlet water monitoring point that is ranked highest and not clustered as the inlet water monitoring point of the current cluster.

[0065] S14. Determine candidate cluster centers based on the number of water inflow monitoring points included in each cluster, the water inflow monitoring point number constraint, and the spatial distance between the water inflow monitoring points of the current cluster and each cluster center. Clustering is performed by selecting the candidate cluster center with the smallest water inflow status correlation coefficient as the cluster center of the water inflow monitoring points of the current cluster. For example, a cluster center including less than 8 water inflow monitoring points and having a spatial distance less than a spatial distance threshold (e.g., 5 kilometers) is selected as a candidate cluster center. If no candidate cluster center exists, the water inflow monitoring point of the current cluster is selected as the new cluster center.

[0066] S15. Update the number of water monitoring points included in each cluster. For example, if the water monitoring points currently clustered are clustered into cluster A, the number of water monitoring points included in cluster A is updated from 5 to 6, and the number of water monitoring points included in other clusters is not updated.

[0067] S16, determine whether all water inlet monitoring points that are not cluster centers have completed clustering, if so, execute S18, if not, execute S17;

[0068] S17. According to the sorting result, the next water inflow monitoring point is used as the water inflow monitoring point of the current cluster;

[0069] S18, determining whether the iteration end condition is met (for example, the number of iterations is greater than a preset number (for example, 50) or the cluster center converges (that is, the number of cluster center changes in multiple consecutive iterations is less than a preset number threshold (for example, 3)), etc.), if so, completing the clustering and outputting multiple clusters, if not, executing S19;

[0070] S19. Update the cluster center according to the number of water inflow monitoring points included in each cluster and the water inflow status correlation coefficient of any two water inflow monitoring points included in each cluster, and execute S12. Specifically, if there is a cluster in which the number of water inflow monitoring points included is less than the minimum number (for example, 3), then the cluster center is discarded. For each cluster whose cluster center is not discarded, for each water inflow monitoring point included in the cluster, the variance can be calculated based on the water inflow status correlation coefficient between the water inflow monitoring point and any other water inflow monitoring point included in the cluster, and the water inflow monitoring point with the largest variance is used as the new cluster center of the cluster.

[0071] For each cluster, a corresponding edge computing node is assigned. This node can be located near the spatial center of the cluster's water monitoring points, thereby determining the optimal edge computing architecture. Each edge computing node is responsible for collecting, processing, and storing data from its corresponding cluster's water monitoring points. For example, real-time analysis of water data is performed to detect anomalies. Based on the number of water monitoring points, data traffic, and computing requirements in each cluster, computing resources (such as CPU, memory, and storage) and communication resources (such as bandwidth) are allocated to the edge computing node. Based on the location of the edge computing nodes, the locations of multiple communication base stations are determined to determine the optimal network architecture. Communication base stations should be placed as close as possible to edge computing nodes to ensure stable data transmission and low latency. A grid or cellular layout can be used, with the location of communication base stations planned based on the density of edge computing nodes and their communication range. A star topology is chosen, with edge computing nodes wirelessly connected to communication base stations, which then transmit data to the water conservancy monitoring platform. Furthermore, a shortest path algorithm is used to optimize data transmission paths and improve network performance.

[0072] Step 140 , for each water inflow monitoring point, determine the heartbeat data packet upload frequency of the water inflow monitoring point based on the multi-source data collected by the multi-source data collection device of the water inflow monitoring point through the optimal edge computing architecture.

[0073] Specifically include:

[0074] Calculate the abnormal water level value of the water inlet monitoring point according to the water level sequence;

[0075] According to the flow velocity sequence, calculate the abnormal flow velocity value of the water inlet monitoring point;

[0076] Calculate the color anomaly value of the water inlet monitoring point based on the image data;

[0077] According to the water level abnormal value, the flow rate abnormal value and the color abnormal value of the water monitoring point, the water abnormal value of the water monitoring point is calculated;

[0078] According to the water abnormal value of the water monitoring point, the heartbeat data packet upload frequency of the water monitoring point is determined.

[0079] Specifically, according to the historical water level data, the upper limit threshold of the water level is set. The mean value of the water level sequence is obtained by averaging the water level sequence, and the deviation degree of the mean value of the water level and the upper limit threshold of the water level is calculated as the water level abnormal value. In combination with the historical flow rate data and the operation rule of the reservoir, the upper limit threshold of the flow rate is set. The mean value of the flow rate sequence is obtained by averaging the flow rate sequence, and the deviation degree of the mean value of the flow rate and the upper limit threshold of the flow rate is calculated as the water level abnormal value.

[0080] The collected image data is preprocessed, including image enhancement, denoising and other operations. For example, the histogram equalization method is used to enhance the image contrast, and the Gaussian filter is used to remove image noise. Color features such as color histogram and color mean value are extracted from the preprocessed image. Taking the color histogram as an example, the color space of the image is divided into several intervals, the number of pixels in each interval is counted, and the color histogram is obtained. The distribution range of the normal color feature is determined by using the historical normal image data. For the current image data, the deviation degree of its color feature and the color feature of the normal image is calculated as the color abnormal value.

[0081] The water level abnormal value, the flow rate abnormal value and the color abnormal value can be weighted and summed to obtain the water abnormal value of the water monitoring point.

[0082] The greater the water abnormal value of the water monitoring point is, the higher the heartbeat data packet upload frequency of the water monitoring point is.

[0083] Step 150, through the optimal network architecture, according to the heartbeat data packet upload frequency of each water monitoring point, the multi-source data collected by the multi-source data acquisition device of the water monitoring point is uploaded to the water conservancy monitoring platform.

[0084] Step 160, the water conservancy monitoring platform fuses the multi-source data collected by the multi-source data acquisition device of the plurality of water monitoring points to predict the reservoir water level.

[0085] Specifically, it includes:

[0086] For each water monitoring point, based on the water state correlation coefficient of any two water monitoring points, the relevant water monitoring point of the water monitoring point is determined, and according to the relevant water monitoring point of the water monitoring point, for example, the water monitoring point with a water state correlation coefficient greater than a water state correlation coefficient threshold (for example, 0.6) is taken as the relevant water monitoring point of the water monitoring point;

[0087] According to the multi-source data collected by the relevant inflow monitoring point and the multi-source data collection device of the plurality of inflow monitoring points of each inflow monitoring point, the water level sequence and the flow rate sequence of the inflow monitoring point are corrected;

[0088] According to the corrected water level sequence and flow rate sequence of the inflow monitoring point, the reservoir water level is predicted.

[0089] Figure 3 is a flow diagram for correcting the water level sequence and the flow rate sequence of the inflow monitoring point according to some embodiments of the present specification, as shown in Figure 3 As preferred, according to the multi-source data collected by the relevant inflow monitoring point and the multi-source data collection device of the plurality of inflow monitoring points of each inflow monitoring point, the water level sequence and the flow rate sequence of the inflow monitoring point are corrected, including:

[0090] S21, the water level sequence and the flow rate sequence of the inflow monitoring point are variational mode decomposed to obtain the intrinsic mode function and the residual of the water level sequence and the intrinsic mode function and the residual of the flow rate sequence;

[0091] S22, for each inflow monitoring point, the intrinsic mode function and the residual of the water level sequence and the intrinsic mode function and the residual of the flow rate sequence of the relevant inflow monitoring point, the intrinsic mode function and the residual of the water level sequence and the intrinsic mode function and the residual of the flow rate sequence of the inflow monitoring point are corrected. Specifically, for each target inflow monitoring point, the intrinsic mode function and the residual of the water level sequence and the intrinsic mode function and the residual of the flow rate sequence of the relevant inflow monitoring point are taken as input, and the intrinsic mode function and the residual of the water level sequence and the intrinsic mode function and the residual of the flow rate sequence of the target inflow monitoring point are taken as output. A correction model is established. The correction model can be a neural network model. The intrinsic mode function and the residual of the water level sequence and the flow rate sequence of the relevant inflow monitoring point are input into the correction model to obtain the correction value of the intrinsic mode function and the residual of the water level sequence and the flow rate sequence of the inflow monitoring point;

[0092] S23, based on the intrinsic mode function and the residual of the water level sequence and the intrinsic mode function and the residual of the flow rate sequence before correction and the intrinsic mode function and the residual of the water level sequence and the intrinsic mode function and the residual of the flow rate sequence after correction of each inflow monitoring point, the global correction change value is calculated. For example, for each inflow monitoring point, the water level sequence and the flow rate sequence can be reconstructed according to the intrinsic mode function and the residual of the water level sequence and the intrinsic mode function and the residual of the flow rate sequence after correction, the cosine distance between the water level sequence before reconstruction and the water level sequence after reconstruction corresponding to the current iteration number and the cosine distance between the flow rate sequence before reconstruction and the flow rate sequence after reconstruction corresponding to the current iteration number are calculated, and the weighted sum is calculated to obtain the correction change value of the inflow monitoring point. The correction change values of each inflow monitoring point are summed to obtain the global correction change value.

[0093] S24, determining whether the global correction change value converges (for example, the difference of the global correction change values ​​of multiple consecutive iterations is less than a preset difference (for example, 0.5)), if so, completing the correction, if not, executing S25;

[0094] S25. For each water inflow monitoring point, data is updated based on the corrected intrinsic mode functions and residuals of the water level sequence and the intrinsic mode functions and residuals of the flow velocity sequence, and S22 is executed. Specifically, for each water inflow monitoring point, the water level sequence and flow velocity sequence can be reconstructed based on the corrected intrinsic mode functions and residuals of the water level sequence and the intrinsic mode functions and residuals of the flow velocity sequence, and the reconstructed water level sequence and flow velocity sequence can be used as the water level sequence and flow velocity sequence before reconstruction in the next iteration.

[0095] The reservoir water level can be predicted based on the water level sequence and flow rate sequence of each water inflow monitoring point and the water level sequence of the reservoir through a water level prediction model, wherein the water level prediction model can be a long short-term memory network model.

[0096] As an option, the water conservancy monitoring platform is also used for:

[0097] Based on the color anomaly values ​​of multiple water inflow monitoring points, the geological disaster warning area is determined.

[0098] Specifically, the location information of multiple water inflow monitoring points and the corresponding color outlier values ​​are spatially associated. Using geographic information system technology, the monitoring points are plotted on the map, and the size of the color outlier values ​​is represented by different colors or symbols. For example, red represents high outlier value areas, yellow represents medium outlier value areas, and green represents low outlier value areas. The color outlier values ​​of the monitoring points are spatially interpolated using spatial interpolation methods (such as Kriging interpolation and inverse distance weighted interpolation) to obtain a color outlier value distribution map for the entire study area. This can provide a more intuitive understanding of the spatial distribution of color anomalies and identify areas that may be at risk of landslides and mudslides (i.e., geological disaster warning areas).

[0099] Figure 4 This is a module diagram of a smart water conservancy monitoring system based on multi-source data fusion according to some embodiments of this specification, such as Figure 4 As shown, the smart water conservancy monitoring system based on multi-source data fusion can include a data acquisition module, a data transmission module and a water level monitoring module.

[0100] The data acquisition module is used to determine multiple water inflow monitoring points of the reservoir based on the historical water inflow data of the reservoir, and set a multi-source data acquisition device at each water inflow monitoring point;

[0101] The data transmission module is configured to determine an optimal edge computing architecture and an optimal network architecture according to a plurality of water inflow monitoring points, for each water inflow monitoring point, determine a heartbeat data packet upload frequency of the water inflow monitoring point according to multi-source data collected by a multi-source data collection device of the water inflow monitoring point through the optimal edge computing architecture, and upload the multi-source data collected by the multi-source data collection device of the water inflow monitoring point to a water conservancy monitoring platform through the optimal network architecture according to the heartbeat data packet upload frequency of each water inflow monitoring point.

[0102] The water level monitoring module is configured to perform reservoir water level prediction by fusing multi-source data collected by a plurality of water inflow monitoring point multi-source data collection devices through the water conservancy monitoring platform.

[0103] The intelligent water conservancy monitoring system based on multi-source data fusion can be used to execute the intelligent water conservancy monitoring method based on multi-source data fusion, which will not be described here.

[0104] Finally, it should be understood that the embodiments described in the specification are only used to illustrate the principles of the embodiments of the specification. Other variations can also belong to the scope of the specification. Therefore, as an example but not limitation, alternative configurations of the embodiments of the specification can be considered consistent with the teachings of the specification. Accordingly, the embodiments of the specification are not limited to the embodiments explicitly introduced and described in the specification.

Claims

1. The intelligent water conservancy monitoring method based on multi-source data fusion is characterized by: include: Determine multiple water inflow monitoring points of the reservoir based on the historical water inflow data of the reservoir; At each water inlet monitoring point, a multi-source data acquisition device is set up; Determine the optimal edge computing architecture and network architecture based on multiple water inflow monitoring points; For each water inflow monitoring point, the optimal edge computing architecture is used to determine the heartbeat data packet upload frequency of the water inflow monitoring point based on the multi-source data collected by the multi-source data acquisition device at the water inflow monitoring point. Through the optimal network architecture, the multi-source data collected by the multi-source data acquisition device at each water monitoring point is uploaded to the water conservancy monitoring platform according to the heartbeat data packet upload frequency of each water monitoring point; The water conservancy monitoring platform integrates multi-source data collected by multi-source data acquisition devices at multiple water inflow monitoring points to predict reservoir water levels; Among them, the optimal edge computing architecture and optimal network architecture are determined based on multiple water monitoring points, including: Determine the water status correlation coefficient of any two water inflow monitoring points based on the historical water inflow data of any two water inflow monitoring points; Clustering multiple inflow water monitoring points is performed based on the water inflow status correlation coefficient and spatial distance between any two inflow water monitoring points, the correlation coefficient between the water inflow status of each inflow water monitoring point and the water level of the reservoir, and the number constraint of inflow water monitoring points to determine multiple clusters; Determine the optimal edge computing architecture and optimal network architecture based on multiple clusters; Based on the water supply status correlation coefficient and spatial distance between any two water supply monitoring points, the correlation coefficient between the water supply status of each water supply monitoring point and the water level of the reservoir, and the number constraint of water supply monitoring points, multiple water supply monitoring points are clustered to determine multiple clusters, including: S11, based on the water status correlation coefficient of any two water monitoring points, select multiple water monitoring points as cluster centers; S12, sorting the water inflow monitoring points that are not the cluster center according to the correlation coefficient between the water inflow status of the water inflow monitoring points and the water level of the reservoir, and generating a sorting result; S13. Determine the water inlet monitoring point of the current cluster according to the sorting result; S14. Determine candidate cluster centers based on the number of water inflow monitoring points included in each cluster, the water inflow monitoring point number constraint, and the spatial distance between the water inflow monitoring points of the current cluster and each cluster center. The candidate cluster center with the smallest water inflow status correlation coefficient is used as the cluster center of the water inflow monitoring points of the current cluster, and clustering is performed. S15, updating the number of water monitoring points included in each cluster; S16, determine whether all water inlet monitoring points that are not cluster centers have completed clustering, if so, execute S18, if not, execute S17; S17. According to the sorting result, the next water inflow monitoring point is used as the water inflow monitoring point of the current cluster; S18, determine whether the iteration end condition is met, if so, complete clustering and output multiple clusters, if not, execute S19; S19. Update the cluster center according to the number of water inflow monitoring points included in each cluster and the water inflow status correlation coefficient between any two water inflow monitoring points included in each cluster, and execute S12.

2. The intelligent water conservancy monitoring method based on multi-source data fusion according to claim 1 is characterized in that: Based on the historical water inflow data of the reservoir, multiple water inflow monitoring points of the reservoir are determined, including: Obtain regional digital elevation models of reservoirs; Determine multiple initial monitoring points based on the regional digital elevation model of the reservoir; Obtain historical water inflow data and historical water level data of multiple initial monitoring points; Determine the correlation coefficient between the water inflow status of the initial monitoring point and the water level of the reservoir based on the historical water inflow data of multiple initial monitoring points and the historical water level data of the reservoir; Through genetic algorithm, multiple water inflow monitoring points of the reservoir are determined according to the correlation coefficient between the water inflow status of each initial monitoring point and the water level of the reservoir.

3. The intelligent water conservancy monitoring method based on multi-source data fusion according to claim 2 is characterized in that: Through genetic algorithms, multiple water inflow monitoring points of the reservoir are determined based on the correlation coefficient between the water inflow status of each initial monitoring point and the water level of the reservoir, including: Based on the historical water inflow data of multiple initial monitoring points, the similarity of the water inflow status and the water inflow status correlation coefficient of the two initial monitoring points are calculated; Sampling is performed from a plurality of initial monitoring points to initialize a plurality of individuals, wherein each individual includes at least three initial monitoring points; Establishing a fitness function, wherein the fitness function is related to the similarity of water inflow states and the water inflow state correlation coefficient of any two initial monitoring points included in the individual, and the correlation coefficient between the water inflow state of any initial monitoring point and the water level of the reservoir; Calculate the fitness value of each individual according to the fitness function; According to the fitness value of each individual, multiple individuals are repeatedly selected, crossed, mutated and terminated until multiple water inflow monitoring points of the reservoir are determined.

4. The intelligent water conservancy monitoring method based on multi-source data fusion according to claim 2 is characterized in that: The multi-source data collected by the multi-source data collection device at the water inlet monitoring point includes at least water level sequence, flow rate sequence and image data; Through the optimal edge computing architecture, the frequency of uploading heartbeat data packets at the water inflow monitoring point is determined based on the multi-source data collected by the multi-source data acquisition device at the water inflow monitoring point, including: Calculate the abnormal water level value of the water inlet monitoring point according to the water level sequence; According to the flow velocity sequence, calculate the abnormal flow velocity value of the water inlet monitoring point; Calculate the color anomaly value of the water inlet monitoring point based on the image data; Calculate the abnormal value of water inflow at the water monitoring point based on the abnormal value of water level, abnormal value of flow velocity and abnormal value of color at the water monitoring point; According to the abnormal value of the water inflow monitoring point, the frequency of uploading the heartbeat data packet of the water inflow monitoring point is determined.

5. The intelligent water conservancy monitoring method based on multi-source data fusion according to any one of claims 1 to 4, characterized in that: The water conservancy monitoring platform integrates multi-source data collected by multi-source data acquisition devices at multiple water inflow monitoring points to predict reservoir water levels, including: For each water inflow monitoring point, based on the water inflow status correlation coefficient of any two water inflow monitoring points, determine the related water inflow monitoring point of the water inflow monitoring point, according to the related water inflow monitoring point of the water inflow monitoring point; Correcting the water level sequence and flow rate sequence of the water inlet monitoring point based on the multi-source data collected by the multi-source data collection device of the relevant water inlet monitoring point and multiple water inlet monitoring points of each water inlet monitoring point; Reservoir water level prediction is carried out based on the revised water level sequence and flow rate sequence of the inflow monitoring points.

6. The intelligent water conservancy monitoring method based on multi-source data fusion according to any one of claims 1 to 4, characterized in that: According to the multi-source data collected by the multi-source data acquisition device of the relevant water inflow monitoring points of each water inflow monitoring point and multiple water inflow monitoring points, the water level sequence and flow rate sequence of the water inflow monitoring point are corrected, including: S21. performing variational mode decomposition on the water level sequence and the flow velocity sequence of the water inlet monitoring point to obtain the intrinsic mode function and residual of the water level sequence and the intrinsic mode function and residual of the flow velocity sequence; S22. For each water inflow monitoring point, the intrinsic mode function and residual of the water level sequence and the intrinsic mode function and residual of the flow velocity sequence of the relevant water inflow monitoring point are corrected; S23, calculating a global correction change value based on the intrinsic mode function and residual of the water level sequence before correction, the intrinsic mode function and residual of the flow velocity sequence, and the intrinsic mode function and residual of the water level sequence after correction, and the intrinsic mode function and residual of the flow velocity sequence of each water inflow monitoring point; S24, determine whether the global correction change value converges, if so, complete the correction, if not, execute S25; S25. For each water inflow monitoring point, update the data based on the corrected intrinsic mode function and residual of the water level sequence and the intrinsic mode function and residual of the flow velocity sequence, and execute S22.

7. The intelligent water conservancy monitoring method based on multi-source data fusion according to any one of claims 1 to 4, characterized in that: The water conservancy monitoring platform is also used for: Based on the color anomaly values ​​of multiple water inflow monitoring points, the geological disaster warning area is determined.

8. The intelligent water conservancy monitoring system based on multi-source data fusion is characterized by: The method for implementing the intelligent water conservancy monitoring method based on multi-source data fusion according to any one of claims 1 to 7 comprises: The data acquisition module is used to determine multiple water inflow monitoring points of the reservoir based on the historical water inflow data of the reservoir, and set a multi-source data acquisition device at each water inflow monitoring point; A data transmission module is used to determine the optimal edge computing architecture and the optimal network architecture based on multiple water monitoring points. For each water monitoring point, the optimal edge computing architecture is used to determine the heartbeat data packet upload frequency of the water monitoring point based on the multi-source data collected by the multi-source data acquisition device at the water monitoring point. The optimal network architecture is used to upload the multi-source data collected by the multi-source data acquisition device at the water monitoring point to the water conservancy monitoring platform based on the heartbeat data packet upload frequency of each water monitoring point. The water level monitoring module is used to predict the reservoir water level by integrating multi-source data collected by multi-source data acquisition devices at multiple water inflow monitoring points through the water conservancy monitoring platform.

Images