Intelligent water conservancy monitoring method and system based on digital twinning

Abstract

The invention discloses an intelligent water conservancy monitoring method and system based on digital twinning, and relates to the field of water conservancy monitoring, and the method comprises the steps: constructing a basic hydrological map structure, generating M meteorological hydrological map structures, sorting the M meteorological hydrological map structures, generating a hydrological map sequence, intercepting Q hydrological map time sequence samples from the hydrological map sequence, the method comprises the following steps: generating a pre-trained water conservancy monitoring model, constructing a meteorological hydrological diagram structure in a current time window, inputting the meteorological hydrological diagram structure into the water conservancy monitoring model to predict reservoir inflow and reservoir outflow of N hydrological nodes in L time windows in the future, and calculating the reservoir inflow and reservoir outflow of the N hydrological nodes in the L time windows in the future based on the reservoir inflow and reservoir outflow of the N hydrological nodes in the L time windows in the future. According to the method, the reservoir entry flow and the reservoir exit flow of the multiple key nodes in the whole watershed in the future time period can be synchronously output, the accuracy of multi-node flow prediction is remarkably improved, and the abnormal water regimen early warning level of water conservancy monitoring is improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of water conservancy monitoring, in particular to a smart water conservancy monitoring method and system based on digital twinning. BACKGROUND

[0002] The current water conservancy monitoring system is mainly used for monitoring the water conservancy operation of rivers, lakes and reservoirs, and timely reflecting the hydrological characteristics of each water area, so that the relevant departments can make arrangements to prevent the occurrence of flood disasters; for example, the monitoring method disclosed in CN115081157A discloses a smart water conservancy monitoring method and system based on the Internet of Things, which can effectively, accurately and intelligently monitor water conservancy. However, as such a water conservancy monitoring method, it usually relies on historical flow data of a single hydrological station for local prediction, and it is difficult to reflect the dynamic correlation and spatial propagation effect between upstream and downstream hydrological nodes. In recent years, although some research has attempted to introduce machine learning methods, the hydrological stations are treated as independent time series, and the topological connection relationship between nodes in the basin and the influence of meteorological driving are ignored, resulting in insufficient collaborative prediction of the inflow / outflow of the whole basin. Multiple nodes, it is difficult to support real-time scheduling and abnormal warning of water conservancy monitoring. SUMMARY

[0003] In view of the deficiencies of the prior art, the present application provides a smart water conservancy monitoring method and system based on digital twinning, which solves the technical problems raised in the background art by introducing a graph structure modeling of the hydrological nodes in the whole basin.

[0004] To achieve the above purpose, the present application realizes the following technical scheme:

[0005] In a first aspect, the present application provides a smart water conservancy monitoring method based on digital twinning, which comprises:

[0006] S1, constructing a basic hydrological graph structure with a target reservoir as a central node within a time window;

[0007] S2, embedding meteorological driving strength into the basic hydrological graph structure in M time windows to generate M meteorological hydrological graph structures;

[0008] S3, sorting the M meteorological hydrological graph structures based on the time sequence of the M time windows to generate a hydrological graph sequence;

[0009] S4, intercepting Q hydrological graph time sequence samples on the hydrological graph sequence;

[0010] S5, sequentially inputting the Q hydrological graph time sequence samples according to the sample training serial number into a graph neural network for supervised training to generate a pre-trained water conservancy monitoring model;

[0011] S6. Construct the meteorological and hydrological map structure within the current time window, and input it into the water conservancy monitoring model to predict the inflow and outflow of N hydrological nodes within the next L time windows;

[0012] S7. Based on the inflow and outflow of N hydrological nodes within the next L time windows, determine the hydrological anomalies of the entire basin.

[0013] In some of these embodiments, a basic hydrological map structure centered on the target reservoir is constructed within a time window, including:

[0014] S1-1. Define the central node of the target reservoir and its N hydrological nodes;

[0015] S1-2. Establish hydrological connection edges between the N hydrological nodes and the central node; wherein, the hydrological connection edges include: upstream connection edges and downstream connection edges;

[0016] S1-3, Calculate the edge weights of each hydrological connection edge;

[0017] S1-4. Assign the edge weights to the corresponding upstream or downstream connecting edges to construct a basic hydrological map structure with the target reservoir as the central node within the time window.

[0018] In some embodiments, a central node of the target reservoir and its N hydrological nodes are defined, including:

[0019] S1-1-1, Anchor the geographical coordinates of the target reservoir on the electronic map;

[0020] S1-1-2. Using the aforementioned geographical coordinates as the center, define the monitoring range of the target reservoir with a preset radius;

[0021] S1-1-3. Within the monitoring range, N hydrological nodes are anchored with the geographical coordinates of the target reservoir as the central node.

[0022] In some embodiments, hydrological connection edges are established between the N hydrological nodes and the central node, including:

[0023] S1-2-1. Select any one of the N hydrological nodes;

[0024] S1-2-2, Obtain the water flow direction between this hydrological node and the central node;

[0025] S1-2-3. If the water flow direction points to the center node, then this hydrological node is determined to be an upstream node, and an upstream connection edge is established.

[0026] S1-2-4. If the water flow direction is away from the central node, then this hydrological node is determined to be a downstream node, and a downstream connection edge is established.

[0027] In some embodiments, the edge weights of each hydrological connection edge are calculated, including:

[0028] S1-3-1. Obtain the geographic coordinates of N hydrological nodes;

[0029] S1-3-2. Calculate the connection distance between each hydrological node and the central node based on the geographical coordinates of each hydrological node and the central node.

[0030] S1-3-3. Divide the time axis into M consecutive time windows with a fixed sampling period;

[0031] S1-3-4. Collect the inflow and outflow of each hydrological node within the specified time window;

[0032] S1-3-5. Based on the inflow and outflow, determine the net water storage of each hydrological node within each time window;

[0033] S1-3-6. Based on the net water storage of each hydrological node within each time window, determine the unit water storage rate of each hydrological node; wherein, the unit water storage rate characterizes the rate of change of the net water storage of the hydrological node within a time window.

[0034] S1-3-7. Based on the unit water storage rate of each hydrological node and its connection distance with the central node, calculate the edge weight of each hydrological connection edge.

[0035] S1-3-8. Assign the edge weights to the corresponding upstream or downstream connecting edges to construct a basic hydrological map structure with the target reservoir as the central node within the time window.

[0036] In some embodiments, M meteorological and hydrological map structures are generated, including:

[0037] S2-1. Obtain the historical rainfall and future forecast rainfall for each hydrological node within the meteorological monitoring range along the start and end times of each time window;

[0038] S2-2. Based on the historical rainfall and the forecasted rainfall, determine the meteorological driving intensity of each hydrological node;

[0039] S2-3. Embed the meteorological driving intensity into the corresponding hydrological nodes in the basic hydrological map structure to generate a meteorological and hydrological map structure with the target reservoir as the central node within the time window.

[0040] S2-4. Traverse M time windows and repeat the generation of meteorological and hydrological map structures until M meteorological and hydrological map structures are generated.

[0041] In some embodiments, Q hydrographic time-series samples are extracted from the hydrographic sequence, including:

[0042] S4-1. Slide and extract Q hydrological map subsequences from the hydrological map sequence; where each hydrological map subsequence contains K frames of meteorological and hydrological map structure corresponding to K time windows;

[0043] S4-2. The last frame meteorological and hydrological map structure of each hydrological map subsequence is used as the supervision target, and the inflow and outflow of each hydrological node contained therein are used as supervision labels to form Q hydrological map samples.

[0044] S4-3. Anchor the K time windows corresponding to each hydrological map sample, and assign sample training numbers to each hydrological map sample based on the K time windows to obtain Q hydrological map time series samples.

[0045] This invention provides a smart water conservancy monitoring method and system based on digital twins, which has the following beneficial effects:

[0046] This invention incorporates multiple hydrological nodes within a watershed (such as river cross-sections, sluice gates, and tributary inlets) into a unified graph structure. Within each time window, it integrates the meteorological and hydrological conditions of the regions where each node is located, constructing a meteorological-hydrological map that reflects actual flow relationships and meteorological impacts. Based on this, a graph neural network is used to learn from the meteorological-hydrological maps across multiple consecutive time windows. In prediction, it not only considers the historical changes of individual nodes but also simultaneously captures the water transfer patterns between upstream and downstream nodes and the differentiated effects of rainfall on different regions. Therefore, it can simultaneously output the inflow and outflow of multiple key nodes across the entire watershed for future periods, significantly improving the accuracy of multi-node flow prediction and enhancing the real-time scheduling and early warning capabilities of water conservancy monitoring.

[0047] Secondly, the present invention provides a smart water conservancy monitoring system based on digital twins, which executes the smart water conservancy monitoring method based on digital twins described in the first aspect, the system comprising:

[0048] The basic hydrological map construction unit is used to construct the basic hydrological map structure with the target reservoir as the central node within a time window.

[0049] The meteorological and hydrological map generation unit is used to embed meteorological driving intensity into the basic hydrological map structure within M time windows to generate M meteorological and hydrological map structures.

[0050] The hydrological map sequence sorting unit is used to sort the structures of M meteorological and hydrological maps based on the time order of M time windows to generate a hydrological map sequence.

[0051] The time series sample acquisition unit is used to extract Q time series samples from the hydrological map sequence.

[0052] The graph time series sample training unit is used to input Q hydrological graph time series samples into the graph neural network in sequence according to the sample training sequence number for supervised training, and generate a pre-trained water conservancy monitoring model.

[0053] The hydrological map structure monitoring unit is used to construct the meteorological and hydrological map structure within the current time window, and input it into the water conservancy monitoring model to predict the inflow and outflow of N hydrological nodes within the next L time windows.

[0054] The hydrological anomaly detection unit is used to detect hydrological anomalies across the entire basin based on the inflow and outflow of N hydrological nodes within L future time windows.

[0055] Compared with the prior art, the beneficial effects of the intelligent water conservancy monitoring system based on digital twins of the present invention are the same as those of the aforementioned intelligent water conservancy monitoring method based on digital twins, and therefore will not be repeated here. Attached Figure Description

[0056] Figure 1 A flowchart illustrating a smart water conservancy monitoring method based on digital twins provided by this invention;

[0057] Figure 2 This is a schematic diagram illustrating the construction process of the basic hydrological map structure described in this invention;

[0058] Figure 3 This is a schematic diagram illustrating the process for determining the edge direction of a hydrological node according to the present invention.

[0059] Figure 4 This is a schematic diagram of the process for assigning edge weights according to the present invention;

[0060] Figure 5 This is a schematic diagram illustrating the process of obtaining hydrological map time-series samples according to the present invention;

[0061] Figure 6 The present invention provides a structural block diagram of a smart water conservancy monitoring system based on digital twins. Detailed Implementation

[0062] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0063] Example 1: Please refer to Figure 1 value Figure 5This invention provides a smart water conservancy monitoring method based on digital twins, comprising the following steps:

[0064] S1. Construct a basic hydrological map structure with the target reservoir as the central node within the time window;

[0065] S2. Within M time windows, embed meteorological driving intensity into the basic hydrological map structure to generate M meteorological and hydrological map structures.

[0066] S3. Sort the M meteorological and hydrological map structures according to the time order of M time windows to generate a hydrological map sequence;

[0067] For example, the center timestamps of M time windows can be selected, and sorting can be performed based on the time order of the center timestamps.

[0068] Specifically, the hydrological map sequence represents a set of M meteorological and hydrological map structures arranged in chronological order, where each map structure corresponds to a time window, and the node attributes and edge attributes in the map structure evolve over time to characterize the dynamic changes in the hydrology of the entire basin over a continuous period of time.

[0069] S4. Extract Q hydrographic time series samples from the hydrographic map sequence;

[0070] S5. Input the Q hydrological map time series samples into the graph neural network in sequence according to the sample training number for supervised training to generate a pre-trained water conservancy monitoring model.

[0071] Specifically, the supervised training-based water conservancy monitoring model refers to a prediction model constructed by end-to-end learning of hydrological time series samples through graph neural networks. Its input is a graph sequence composed of K frames of continuous meteorological and hydrological graph structures, and its output is the inflow and outflow of each hydrological node corresponding to the last frame of the meteorological and hydrological graph structure in the graph sequence. The model parameters are iteratively optimized by minimizing the loss function (such as mean square error) between the predicted value and the true label until convergence.

[0072] Furthermore, the hydrological map time series samples used as training samples are modeled on the time axis with the temporal dependencies between multiple samples. The dynamic changes in hydrological state within a continuous time window are captured by the temporal coding module of the graph neural network (such as GRU, LSTM or Temporal GNN). As for the hydrological connection relationship between each hydrological node in the whole basin, the propagation path and mutual influence of water flow are represented by the message passing mechanism of the graph structure, thereby realizing the coordinated expression of the basin hydrological system in the spatiotemporal dimension.

[0073] S6. Construct the meteorological and hydrological map structure within the current time window, and input it into the water conservancy monitoring model to predict the inflow and outflow of N hydrological nodes within the next L time windows;

[0074] Specifically, the meteorological and hydrological map structure within the current time window is constructed based on a real-time mirror of the water area digital twin system. This water area digital twin system is a virtual simulation environment operating in a spatiotemporally synchronized manner. Its state is driven by real-time feedback from various sensors, including the dynamic attributes of each hydrological node (such as flow rate and water level), meteorological field distribution, and network topology, forming a digital model highly consistent with the real watershed.

[0075] S7. Based on the inflow and outflow of N hydrological nodes within the next L time windows, determine the hydrological anomalies of the entire basin.

[0076] It should be noted that the aforementioned determination of hydrological anomalies across the entire basin refers to the following: after obtaining the predicted inflow and outflow of each hydrological node within the next L time windows, based on predefined hydrological data rules, constructing or extracting key features of each hydrological node within each time window, such as flow mutation rate, net water storage deviation, and upstream-downstream flow imbalance index; subsequently, by setting thresholds or using a rule engine, comparing and analyzing these features to determine whether there are any abnormal situations such as exceeding limits, abnormal trends, or logical contradictions, thereby achieving dynamic anomaly warning for the hydrological status of the entire basin.

[0077] For example, when the predicted inflow of a certain upstream river section node increases sharply (e.g., the increase exceeds 50%) within two consecutive time windows, while the outflow of its downstream sluice gate does not increase synchronously, the system can determine it as "abnormal upstream water inflow" or "downstream regulation lag". If the net water storage of the node is continuously negative and the absolute value is more than twice the historical average, it may indicate illegal water extraction or measurement equipment failure. Once such an anomaly triggers the preset rules, alarm information can be generated and pushed to the operation and maintenance platform to assist management personnel in responding quickly.

[0078] Furthermore, hydrological anomaly determination can be based on the following unified mathematical expression: For any hydrological node v within the t-th time window, if it outputs the inflow and outflow of this hydrological node, then the corresponding hydrological rise rate is:

[0079] ;

[0080] in, Indicates the rate of water level rise. Indicates the amount of water entering the reservoir. Indicates the amount of water discharged from the reservoir. This indicates the effective water surface area of ​​this hydrological node. Indicates the length of the time window.

[0081] If the rate of water level rise exceeds the preset water level change rate threshold within multiple consecutive time windows, the hydrological node is determined to be abnormal, and corresponding hydrological interventions are implemented.

[0082] In this embodiment, by constructing a time-series meteorological and hydrological map structure centered on the target reservoir and learning its dynamic change patterns based on a graph neural network, the system achieves monitoring of future inflow and outflow flows of hydrological nodes across the entire basin. Furthermore, by combining hydrological rules, the system identifies potential abnormal states, enabling the digital twin water conservancy system to have the monitoring capability from hydrological state perception to anomaly prediction.

[0083] Specifically, in this embodiment, step S1 includes:

[0084] S1-1. Define the central node of the target reservoir and its N hydrological nodes;

[0085] S1-2. Establish hydrological connection edges between the N hydrological nodes and the central node; wherein, the hydrological connection edges include: upstream connection edges and downstream connection edges;

[0086] Specifically, the hydrological connection edge represents the hydrological connection relationship established between the central node and any hydrological node based on water flow connectivity.

[0087] S1-3, Calculate the edge weights of each hydrological connection edge;

[0088] S1-4. Assign the edge weights to the corresponding upstream or downstream connecting edges to construct a basic hydrological map structure with the target reservoir as the central node within the time window.

[0089] In this embodiment, the basic hydrological map structure is constructed with the target reservoir as the central node. By distinguishing between upstream and downstream connecting edges, the direction of water flow between each hydrological node and the central node is clarified. Furthermore, by assigning edge weights, the connection strength is quantified, thereby expressing the hydrological connection relationship of the local watershed at the graph structure level.

[0090] In this embodiment, step S1-1 further includes:

[0091] S1-1-1, Anchor the geographical coordinates of the target reservoir on the electronic map;

[0092] Specifically, the electronic map is a digital map containing the geographical coordinates of water conservancy facilities.

[0093] S1-1-2. Using the aforementioned geographical coordinates as the center, define the monitoring range of the target reservoir with a preset radius;

[0094] S1-1-3. Within the monitoring range, N hydrological nodes are anchored with the geographical coordinates of the target reservoir as the central node.

[0095] Specifically, the hydrological nodes are water conservancy facilities with flow monitoring capabilities, such as upstream river sections, sluice gates, and tributary confluences.

[0096] In this embodiment, by anchoring the geographical coordinates of the target reservoir on the electronic map and delineating the monitoring range, the N hydrological nodes determined are all located within the effective influence area around the target reservoir and have flow monitoring capabilities, ensuring that the data source of the hydrological map structure has spatial rationality.

[0097] In this embodiment, step S1-2 further includes:

[0098] S1-2-1. Select any one of the N hydrological nodes;

[0099] S1-2-2, Obtain the water flow direction between this hydrological node and the central node;

[0100] S1-2-3. If the water flow direction points to the center node, then this hydrological node is determined to be an upstream node, and an upstream connection edge is established.

[0101] S1-2-4. If the water flow direction is away from the central node, then this hydrological node is determined to be a downstream node, and a downstream connection edge is established.

[0102] In this embodiment, upstream and downstream nodes are distinguished according to the actual water flow direction, and corresponding connection edges are established respectively, so as to accurately reflect the convergence and diversion paths of water flow in the basin in the graph structure.

[0103] In this embodiment, step S1-3 further includes:

[0104] S1-3-1. Obtain the geographic coordinates of N hydrological nodes;

[0105] S1-3-2. Calculate the connection distance between each hydrological node and the central node based on the geographical coordinates of each hydrological node and the central node.

[0106] For example, the geographic coordinates are preferably latitude and longitude coordinates in the WGS84 coordinate system, and the connection distance is calculated using the great circle distance formula.

[0107] S1-3-3. Divide the time axis into M consecutive time windows with a fixed sampling period;

[0108] S1-3-4. Collect the inflow and outflow of each hydrological node within the specified time window;

[0109] S1-3-5. Based on the inflow and outflow, determine the net water storage of each hydrological node within each time window;

[0110] S1-3-6. Based on the net water storage of each hydrological node within each time window, determine the unit water storage rate of each hydrological node; wherein, the unit water storage rate characterizes the rate of change of the net water storage of the hydrological node within a time window.

[0111] S1-3-7. Based on the unit water storage rate of each hydrological node and its connection distance with the central node, calculate the edge weight of each hydrological connection edge.

[0112] For example, the formula for calculating the edge weight is:

[0113] ;

[0114] in:

[0115] Indicates hydrological nodes The edge weight of the hydrological connection edge between the node and the central node is used to characterize the intensity of the hydrological influence of the node on the central reservoir.

[0116] Indicates hydrological nodes The unit water storage rate within a time window is equal to the net water storage volume divided by the length of the time window.

[0117] It represents the absolute value of the unit water storage rate, which reflects the activity level of the node in water exchange. The larger the value, the greater the water volume change that occurs at the node per unit time.

[0118] Indicates hydrological nodes The connection distance to the central node;

[0119] This represents the preset distance attenuation coefficient, used to characterize the reduction effect of the impact caused by the loss of water flow along the flow path, and can be set based on the empirical characteristics of the watershed.

[0120] This represents the distance transmission loss, indicating that as the connection distance increases, the effective influence of the hydrological node on the central reservoir decreases exponentially.

[0121] Specifically, the edge weight of a hydrological connection reflects the following two key factors:

[0122] ① The level of activity in water exchange, from The decision reflects the principle that "the higher the rate, the greater the change in water volume"; that is, within a fixed time window, the higher the unit water storage rate, the greater the actual scale of water exchange that node participates in during that period.

[0123] ②Distance transmission loss, due to The decision reflects the principle that "the shorter the distance, the stronger the impact." Even if a node is very active, if it is far from the central reservoir, its actual contribution to the central reservoir will be significantly reduced due to factors such as river seepage, evaporation, or water intake.

[0124] In other words, a hydrological node will have a higher edge weight only when it has both a high unit water storage rate (i.e., high activity level) and is close to the central reservoir (i.e., low transmission loss), thus being given a stronger connection weight in the graph structure.

[0125] S1-3-8. Assign the edge weights to the corresponding upstream or downstream connecting edges to construct a basic hydrological map structure with the target reservoir as the central node within the time window.

[0126] It should be noted that this hydrological map structure is built using data from a specific time interval. It displays the hydrological connections between various hydrological nodes (such as river cross-sections, sluice gates, and tributary inlets) within that time period. Each hydrological node is defined as a node in the graph, and the connections between nodes represent the direction and intensity of water flow, i.e., edge weights. Through this construction method, a new hydrological map is generated after each time window, thus providing a better understanding of the hydrological conditions at different times.

[0127] Specifically, in this embodiment, the meteorological and hydrological map structure is the data carrier of the target reservoir and its associated watershed in the digital twin system.

[0128] In this embodiment, the edge weights are calculated by combining the unit water storage rate of the hydrological node with the connection distance to the central node. This allows the weight of each hydrological connection edge to simultaneously reflect the activity level of water exchange at the node and the distance transmission attenuation effect during water flow propagation, thereby realizing a quantitative expression of the intensity of hydrological influence in the graph structure.

[0129] Specifically, in this embodiment, step S2 includes:

[0130] S2-1. Obtain the historical rainfall and future forecast rainfall for each hydrological node within the meteorological monitoring range along the start and end times of each time window;

[0131] It should be noted that the meteorological monitoring range refers to a circular area with a radius of R centered on the geographical coordinates of each hydrological node; future rainfall forecasts can be obtained from data released by the relevant meteorological bureaus.

[0132] S2-2. Based on the historical rainfall and the forecasted rainfall, determine the meteorological driving intensity of each hydrological node;

[0133] For example, the meteorological driving intensity is defined as the weighted sum of historical rainfall and future forecast rainfall.

[0134] S2-3. Embed the meteorological driving intensity into the corresponding hydrological nodes in the basic hydrological map structure to generate a meteorological and hydrological map structure with the target reservoir as the central node within the time window.

[0135] Specifically, the meteorological and hydrological map structure represents an enhanced graph that embeds meteorological driving intensity and hydrological connectivity (edge ​​weights) within a time window. The node attributes of each hydrological node include unit water storage rate and meteorological driving intensity, and the edge attributes are the edge weights of the hydrological connectivity edges.

[0136] S2-4. Traverse M time windows and repeat the generation of meteorological and hydrological map structures until M meteorological and hydrological map structures are generated.

[0137] In this embodiment, by embedding meteorological driving intensity as a node attribute into the basic hydrological map structure, the generated meteorological-hydrological map structure can simultaneously reflect the unit water storage rate and meteorological influences of each hydrological node within the watershed. This process is repeated over M consecutive time windows, ultimately forming a series of meteorological-hydrological map structures that dynamically display changes in hydrological status and their response to meteorological factors.

[0138] Specifically, in this embodiment, step S4 includes:

[0139] S4-1. Slide and extract Q hydrological map subsequences from the hydrological map sequence; where each hydrological map subsequence contains K frames of meteorological and hydrological map structure corresponding to K time windows;

[0140] S4-2. The last frame meteorological and hydrological map structure of each hydrological map subsequence is used as the supervision target, and the inflow and outflow of each hydrological node contained therein are used as supervision labels to form Q hydrological map samples.

[0141] In other words, the hydrological map sample represents a supervised learning sample composed of an input sequence consisting of a K-frame continuous meteorological and hydrological map structure, and a supervised learning sample composed of the inflow and outflow of each hydrological node in the last frame of the sequence.

[0142] S4-3. Anchor the K time windows corresponding to each hydrological map sample, and assign sample training sequence numbers to each hydrological map sample based on the K time windows to obtain Q hydrological map time series samples.

[0143] For example, the earliest or latest timestamp within the K time windows covered by each hydrographic map sample can be used as its unique identifier, and all hydrographic map samples can be sorted based on the order of these timestamps, thereby assigning a unique sample training sequence number to each sample.

[0144] In this embodiment, by sliding and cropping K consecutive frames of meteorological and hydrological map structure on the hydrological map sequence as input, and using the inflow and outflow of the last frame as supervision labels, the constructed hydrological map time series sample naturally has time series dependence and clear supervision signals, thereby being able to fit the dynamic change process of watershed hydrological status from meteorological and hydrological to flow results.

[0145] Example 2: This Example 2 differs from Example 1 in that it also provides a smart water conservancy monitoring system based on digital twins. This system is used to implement the above-described method embodiments, and details already described will not be repeated. The terms "module," "unit," and "subunit" used below refer to combinations of software and / or hardware that achieve a predetermined function. Although the system described in the following embodiments is preferably implemented in software, hardware implementation, or a combination of software and hardware, is also possible and contemplated.

[0146] Figure 6 This is a structural block diagram of a smart water conservancy monitoring system based on digital twins according to the present invention. The system includes:

[0147] The basic hydrological map construction unit is used to construct the basic hydrological map structure with the target reservoir as the central node within a time window.

[0148] The meteorological and hydrological map generation unit is used to embed meteorological driving intensity into the basic hydrological map structure within M time windows to generate M meteorological and hydrological map structures.

[0149] The hydrological map sequence sorting unit is used to sort the structures of M meteorological and hydrological maps based on the time order of M time windows to generate a hydrological map sequence.

[0150] The time series sample acquisition unit is used to extract Q time series samples from the hydrological map sequence.

[0151] The graph time series sample training unit is used to input Q hydrological graph time series samples into the graph neural network in sequence according to the sample training sequence number for supervised training, and generate a pre-trained water conservancy monitoring model.

[0152] The hydrological map structure monitoring unit is used to construct the meteorological and hydrological map structure within the current time window, and input it into the water conservancy monitoring model to predict the inflow and outflow of N hydrological nodes within the next L time windows.

[0153] The hydrological anomaly detection unit is used to detect hydrological anomalies across the entire basin based on the inflow and outflow of N hydrological nodes within L future time windows.

[0154] In the aforementioned system, a basic hydrological map structure is constructed through a basic hydrological map construction unit; M meteorological and hydrological map structures are generated through a meteorological and hydrological map generation unit; a hydrological map sequence is generated through a hydrological map sequence sorting unit; Q hydrological map time series samples are obtained through a map time series sample acquisition unit; a pre-trained water conservancy monitoring model is generated through a map time series sample training unit; the inflow and outflow of N hydrological nodes within the next L time windows are predicted through a hydrological map structure monitoring unit; and hydrological anomaly determination is performed for the entire basin through a hydrological map anomaly determination unit, thus solving the problem of insufficient collaborative prediction.

[0155] The above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented, in whole or in part, as a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer program are loaded or executed on a computer, all or part of the processes or functions described in the embodiments of this application are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired (e.g., infrared, wireless, microwave, etc.) means.

[0156] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application.

Claims

1. A smart water conservancy monitoring method based on digital twins, characterized in that, include: S1. Construct a basic hydrological map structure with the target reservoir as the central node within the time window; S2. Within M time windows, embed meteorological driving intensity into the basic hydrological map structure to generate M meteorological and hydrological map structures. S3. Sort the M meteorological and hydrological map structures according to the time order of M time windows to generate a hydrological map sequence; S4. Extract Q hydrographic time series samples from the hydrographic map sequence; S5. Input the Q hydrological map time series samples into the graph neural network in sequence according to the sample training number for supervised training to generate a pre-trained water conservancy monitoring model. S6. Construct the meteorological and hydrological map structure within the current time window, and input it into the water conservancy monitoring model to predict the inflow and outflow of N hydrological nodes within the next L time windows; S7. Based on the inflow and outflow of N hydrological nodes within the next L time windows, determine the hydrological anomalies of the entire basin.

2. The intelligent water conservancy monitoring method based on digital twins according to claim 1, characterized in that, Construct a basic hydrological map structure with the target reservoir as the central node within the time window, including: S1-1. Define the central node of the target reservoir and its N hydrological nodes; S1-2. Establish hydrological connection edges between the N hydrological nodes and the central node; wherein, the hydrological connection edges include: upstream connection edges and downstream connection edges; S1-3, Calculate the edge weights of each hydrological connection edge; S1-4. Assign the edge weights to the corresponding upstream or downstream connecting edges to construct a basic hydrological map structure with the target reservoir as the central node within the time window.

3. The intelligent water conservancy monitoring method based on digital twins according to claim 2, characterized in that, Define the central node of the target reservoir and its N hydrological nodes, including: S1-1-1, Anchor the geographical coordinates of the target reservoir on the electronic map; S1-1-2. Using the aforementioned geographical coordinates as the center, define the monitoring range of the target reservoir with a preset radius; S1-1-3. Within the monitoring range, N hydrological nodes are anchored with the geographical coordinates of the target reservoir as the central node.

4. The intelligent water conservancy monitoring method based on digital twin according to claim 2, characterized in that, Establish hydrological connection edges between the N hydrological nodes and the central node, including: S1-2-1. Select any one of the N hydrological nodes; S1-2-2, Obtain the water flow direction between this hydrological node and the central node; S1-2-3. If the water flow direction points to the center node, then this hydrological node is determined to be an upstream node, and an upstream connection edge is established. S1-2-4. If the water flow direction is away from the central node, then this hydrological node is determined to be a downstream node, and a downstream connection edge is established.

5. The intelligent water conservancy monitoring method based on digital twin according to claim 2, characterized in that, Calculate the edge weights of each hydrological connection edge, including: S1-3-1. Obtain the geographic coordinates of N hydrological nodes; S1-3-2. Calculate the connection distance between each hydrological node and the central node based on the geographical coordinates of each hydrological node and the central node. S1-3-3. Divide the time axis into M consecutive time windows with a fixed sampling period; S1-3-4. Collect the inflow and outflow of each hydrological node within the specified time window; S1-3-5. Based on the inflow and outflow, determine the net water storage of each hydrological node within each time window; S1-3-6. Based on the net water storage of each hydrological node within each time window, determine the unit water storage rate of each hydrological node; wherein, the unit water storage rate characterizes the rate of change of the net water storage of the hydrological node within a time window. S1-3-7. Based on the unit water storage rate of each hydrological node and its connection distance with the central node, calculate the edge weight of each hydrological connection edge. S1-3-8. Assign the edge weights to the corresponding upstream or downstream connecting edges to construct a basic hydrological map structure with the target reservoir as the central node within the time window.

6. The intelligent water conservancy monitoring method based on digital twin according to claim 1, characterized in that, Within M time windows, meteorological driving forces are embedded into the basic hydrological map structure to generate M meteorological and hydrological map structures, including: S2-1. Obtain the historical rainfall and future forecast rainfall for each hydrological node within the meteorological monitoring range along the start and end times of each time window; S2-2. Based on the historical rainfall and the forecasted rainfall, determine the meteorological driving intensity of each hydrological node; S2-3. Embed the meteorological driving intensity into the corresponding hydrological nodes in the basic hydrological map structure to generate a meteorological and hydrological map structure with the target reservoir as the central node within the time window. S2-4. Traverse M time windows and repeat the generation of meteorological and hydrological map structures until M meteorological and hydrological map structures are generated.

7. A smart water conservancy monitoring method based on digital twins according to claim 6, characterized in that, Extract Q hydrographic time series samples from the hydrographic map sequence, including: S4-1. Slide and extract Q hydrological map subsequences from the hydrological map sequence; where each hydrological map subsequence contains K frames of meteorological and hydrological map structure corresponding to K time windows; S4-2. The last frame meteorological and hydrological map structure of each hydrological map subsequence is used as the supervision target, and the inflow and outflow of each hydrological node contained therein are used as supervision labels to form Q hydrological map samples. S4-3. Anchor the K time windows corresponding to each hydrological map sample, and assign sample training numbers to each hydrological map sample based on the K time windows to obtain Q hydrological map time series samples.

8. A smart water conservancy monitoring system based on digital twins, comprising the smart water conservancy monitoring method based on digital twins as described in any one of claims 1-7, characterized in that, include: The basic hydrological map construction unit is used to construct the basic hydrological map structure with the target reservoir as the central node within a time window. The meteorological and hydrological map generation unit is used to embed meteorological driving intensity into the basic hydrological map structure within M time windows to generate M meteorological and hydrological map structures. The hydrological map sequence sorting unit is used to sort the structures of M meteorological and hydrological maps based on the time order of M time windows to generate a hydrological map sequence. The time series sample acquisition unit is used to extract Q time series samples from the hydrological map sequence. The graph time series sample training unit is used to input Q hydrological graph time series samples into the graph neural network in sequence according to the sample training sequence number for supervised training, and generate a pre-trained water conservancy monitoring model. The hydrological map structure monitoring unit is used to construct the meteorological and hydrological map structure within the current time window, and input it into the water conservancy monitoring model to predict the inflow and outflow of N hydrological nodes within the next L time windows. The hydrological anomaly detection unit is used to detect hydrological anomalies across the entire basin based on the inflow and outflow of N hydrological nodes within L future time windows.

Images