A hydrological real-time monitoring system

Abstract

This invention relates to the field of hydrological monitoring technology and discloses a real-time hydrological monitoring system. The system includes a hydrological data access module, a hydrological topology modeling module, a confidence quantification module, an anomaly pattern recognition module, a cross-validation arbitration module, and an early warning decision support module. The system receives heterogeneous hydrological data streams through a multi-source sensor adaptation unit, constructs a dynamic hydrological spatiotemporal adjacency graph through spatiotemporal alignment and standardization, quantifies the joint confidence distribution of nodes and edges using neighborhood consistency analysis and Bayesian fusion, identifies suspected hydrological anomalies through topological template matching and confidence threshold screening, performs weighted arbitration based on redundant data source calls and expert rule reasoning to generate a final verification conclusion, and finally outputs scheduling suggestions by combining flood and drought risk quantification and contingency plan matching. This invention ensures the authenticity and reliability of the dynamic hydrological spatiotemporal adjacency graph and improves the reliability and scientific nature of early warning decisions.

Description

Technical Field

[0001] This invention relates to the field of hydrological monitoring technology, specifically to a real-time hydrological monitoring system. Background Technology

[0002] The hydrological monitoring system is suitable for hydrological departments to monitor hydrological parameters such as rivers, lakes, reservoirs, canals and groundwater in real time. The monitoring content includes: water level, flow rate, flow velocity, rainfall, evaporation, sediment, ice jams, soil moisture, water quality, etc.

[0003] Currently, due to the dispersed deployment of watershed hydrological monitoring stations and the variety of equipment types, when real-time access to hydrological data is carried out, the equipped multi-source sensor adapter unit can receive heterogeneous data streams, but it cannot verify in real time the data drift caused by environmental interference and equipment aging of each sensor. When there are implicit deviations in the original monitoring values, it will cause the accumulation of errors in subsequent topology modeling and confidence quantification, and cannot guarantee the accuracy of anomaly detection.

[0004] Therefore, a real-time hydrological monitoring system is proposed to solve the above problems. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this invention provides a real-time hydrological monitoring system that solves the problems mentioned in the background section, such as the inability to verify data drift caused by environmental interference and equipment aging of various sensors in real time, and the inability to guarantee the accuracy of anomaly detection.

[0006] To achieve the above objectives, the present invention provides the following technical solution: a real-time hydrological monitoring system, comprising:

[0007] The hydrological data access module uses a multi-source sensor adaptation unit to receive heterogeneous hydrological data streams, performs timestamp synchronization and spatial coordinate mapping through a spatiotemporal alignment unit, and outputs formatted hydrological data through a data standardization unit.

[0008] The hydrological topology modeling module receives the formatted hydrological data, identifies key hydrological monitoring entities through the monitoring node extraction unit, generates edge sets based on hydrological physical constraints through the dynamic adjacency calculation unit, and outputs a dynamic hydrological spatiotemporal adjacency graph through the graph structure update unit.

[0009] The confidence quantification module receives the dynamic hydrological spatiotemporal adjacency graph, calculates the spatial correlation of node data through the neighborhood consistency analysis unit, compares the hydrological evolution pattern through the hydrological prior verification unit, and outputs the joint confidence distribution of nodes and edges through the Bayesian fusion unit.

[0010] The anomaly pattern recognition module receives the joint confidence distribution and formatted hydrological data, identifies the preset hydrological anomaly map structure through the topology template matching unit, extracts low-confidence node clusters through the confidence threshold filtering unit, and outputs suspected hydrological anomaly events and propagation paths through the spatiotemporal evolution tracking unit.

[0011] The cross-validation arbitration module receives the suspected hydrological anomaly event, obtains independent monitoring data through the redundant data source calling unit, performs domain logic judgment using the hydrological rule reasoning unit, and generates the final verification conclusion and anomaly confidence level through the weighted arbitration unit.

[0012] The early warning decision support module receives the final verification conclusion, calculates the potential impact range through the flood and drought risk quantification unit, generates scheduling suggestions by querying the disposal strategy library through the contingency plan matching unit, and feeds them back to the front-end hydrological monitoring platform through the instruction distribution unit.

[0013] Preferably, the process in the hydrological data access module of receiving heterogeneous hydrological data streams using a multi-source sensor adaptation unit, performing timestamp synchronization and spatial coordinate mapping through a spatiotemporal alignment unit, and outputting formatted hydrological data through a data standardization unit is as follows:

[0014] The multi-source sensor adapter unit is configured to be compatible with heterogeneous data protocols of rain gauges, water level gauges, flow meters, water quality sensors and mobile monitoring buoys, and to uniformly decode pulse, analog and digital signals;

[0015] The spatiotemporal alignment unit is configured to align asynchronous hydrological data points collected by different sensors to a standard time slice based on a unified hydrological time base and time window, and to assign unique geographic coordinates and watershed logical partition identifiers to each data source according to the preset watershed digital elevation model and hydrological station network topology.

[0016] The data standardization unit configuration is used to convert aligned hydrological data into a standardized data structure that includes physical dimensions, monitoring element types, spatial location codes, and timestamps.

[0017] Preferably, the process in the hydrological topology modeling module of receiving the formatted hydrological data, identifying key hydrological monitoring entities through the monitoring node extraction unit, generating edge sets based on hydrological physical constraints through the dynamic adjacency calculation unit, and outputting a dynamic hydrological spatiotemporal adjacency graph through the graph structure update unit is as follows:

[0018] The monitoring node extraction unit is configured to parse hydrological monitoring stations, reservoir gates, dangerous sections of dikes, and hydrological characteristic sections from formatted hydrological data as graph nodes, and extract the real-time hydrological element values ​​of the nodes;

[0019] The dynamic adjacency calculation unit is configured to define hydrological spatial adjacency rules and hydrological temporal adjacency rules. The spatial adjacency rules determine the physical correlation between nodes based on the connectivity of the watershed system, the direction of the river channel, and a preset distance threshold. The temporal adjacency rules determine the temporal correlation of data between nodes based on the propagation lag characteristics of the hydrological process line, and calculate the comprehensive adjacency strength by fusing the spatial distance attenuation factor and the temporal lag factor.

[0020] The graph structure update unit is configured to establish a weighted hydrological spatiotemporal adjacency edge between two nodes when the overall adjacency strength exceeds a preset connection threshold. The weight value is proportional to the overall adjacency strength, and the addition, deletion, and weight update of the graph structure are dynamically maintained as new data is input.

[0021] Preferably, the confidence quantification module receives the dynamic hydrological spatiotemporal adjacency graph, calculates the spatial correlation of node data through the neighborhood consistency analysis unit, compares the hydrological evolution patterns using the hydrological prior verification unit, and outputs the joint confidence distribution of nodes and edges through the Bayesian fusion unit as follows:

[0022] The neighborhood consistency analysis unit is configured to extract the hydrological data of all upstream and downstream adjacent nodes connected by hydrological spatiotemporal adjacency edges within the same time window for the target hydrological node, calculate the similarity between the target node data and the data of each adjacent node, and generate an internal consistency score that represents the consistency between the node data and its hydrological neighborhood after aggregation and normalization.

[0023] The hydrological prior verification unit is configured to compare the current hydrological data of the target node with the theoretical value range calculated based on historical hydrological characteristic values, equipment calibration error range and water balance physical equation, and output the prior conformity score.

[0024] The Bayesian fusion unit is configured to construct a Bayesian network with the hydrological state of the nodes as the latent variable. The internal consistency score, prior conformity score and data quality index are used as the input of observation evidence to infer the joint probability distribution of the nodes in normal, suspicious and abnormal states, and the confidence of the hydrological association represented by the hydrological spatiotemporal adjacency edge is derived from this.

[0025] Preferably, the process by which the anomaly pattern recognition module receives the joint confidence distribution and formatted hydrological data, identifies the preset hydrological anomaly map structure through the topology template matching unit, extracts low-confidence node clusters using the confidence threshold filtering unit, and outputs suspected hydrological anomaly events and propagation paths through the spatiotemporal evolution tracking unit is as follows:

[0026] The topology template matching unit is configured with a predefined hydrological anomaly map topology template that includes flash floods, pollutant plume diffusion, waterlogging circles, and dam break chain transmission. The dynamic hydrological spatiotemporal adjacency graph is matched with the template to identify candidate anomaly clusters that conform to the characteristics of anomaly spatial distribution.

[0027] The confidence threshold filtering unit is configured to set a lower limit for node confidence and an upper limit for edge confidence, and to perform two-layer filtering on candidate abnormal clusters, removing nodes with low self-confidence and high confidence in their associations, and retaining low-confidence nodes with abnormal characteristics and questionable data reliability.

[0028] The spatiotemporal evolution tracking unit is configured to perform spatiotemporal clustering on the selected low-confidence nodes, trace the source of the anomaly and predict the scope of impact based on the direction and speed of hydrological propagation, and generate suspected hydrological anomaly events that include core anomaly nodes, affected watershed zoning and evolution trends.

[0029] Preferably, the process by which the cross-validation arbitration module receives the suspected hydrological anomaly, obtains independent monitoring data through the redundant data source calling unit, performs domain logic judgment using the hydrological rule reasoning unit, and generates the final verification conclusion and anomaly confidence level through the weighted arbitration unit is as follows:

[0030] The redundant data source calling unit is configured to parse the watershed range and element type involved in suspected hydrological anomalies, and retrieve independent third-party monitoring data for comparison from the preset redundant data source registry containing satellite remote sensing inversion data, meteorological radar rainfall data and manual patrol reporting data;

[0031] The hydrological rule reasoning unit is equipped with an expert rule base containing empirical formulas for rainstorm and flood forecasting, water level-discharge relationship curves, and flood control and drought relief dispatch procedures. It matches the characteristic parameters of suspected hydrological anomalies with the conditions in the rule base, performs logical deduction, and outputs verification opinions and weights based on domain knowledge.

[0032] The weighted arbitration unit is configured to comprehensively consider the anomaly degree of the main monitoring data, the consistency results of redundant data, and the expert rule reasoning opinions. It eliminates misjudgments from a single source of information through a weighted voting mechanism and outputs a final verification conclusion with a clear confidence level.

[0033] Preferably, the process by which the early warning decision support module receives the final verification conclusion, calculates the potential impact range through the flood and drought risk quantification unit, queries the response strategy library using the contingency plan matching unit to generate scheduling suggestions, and feeds them back to the front-end hydrological monitoring platform through the instruction distribution unit is as follows:

[0034] The flood and drought risk quantification unit is configured to combine the underlying surface conditions of the watershed, the distribution of socio-economic assets, and the abnormal confidence level in the final verification conclusion to calculate the inundation range, water depth, and duration of flood and drought disasters, and to assess the risk level.

[0035] The contingency plan matching unit is configured to use the anomaly type, risk level, and protected objects within the scope of impact as indexes to query the flood control contingency plan database and match the corresponding engineering dispatch instructions, mass evacuation routes, and material allocation plans.

[0036] The instruction distribution unit is configured to encapsulate the final verification conclusions, risk quantification assessment reports, and disposal recommendations into standardized decision instruction packages, which are then distributed to monitoring and early warning platforms, reservoir dispatch centers, and emergency command terminals in relevant regions.

[0037] Preferably, the process by which the dynamic adjacency calculation unit generates edge sets based on hydrophysical constraints is further as follows:

[0038] The dynamic adjacency calculation unit is configured to incorporate river confluence time parameters and energy attenuation coefficients when calculating the comprehensive adjacency strength between any two hydrological monitoring nodes. The spatial adjacency strength value is calculated as a negative exponential function of the actual flow distance between nodes along the main channel. The temporal adjacency strength value is calculated as a negative exponential function of the difference between the peak monitoring times between nodes and the deviation of the confluence time. The comprehensive adjacency strength is a weighted sum of spatial and temporal adjacency strengths, and the weighting coefficient is dynamically adjusted with the rainfall intensity in the basin.

[0039] Preferably, the process by which the neighborhood consistency analysis unit calculates the spatial correlation of node data further comprises:

[0040] The neighborhood consistency analysis unit is configured to automatically switch similarity algorithms according to the type of hydrological element when calculating the data similarity between the target node and its neighboring nodes. For water level and flow time series, dynamic time-normalized distance reciprocal is used; for spatial distribution of rainfall, Kriging interpolation residual is used; and for water quality concentration, relative deviation reciprocal is used. The obtained multiple sets of similarity values ​​are then weighted and averaged according to the importance of the watershed location of the neighboring nodes to generate an internal consistency score.

[0041] Preferably, the process by which the weighted arbitration unit generates the final verification conclusion further comprises:

[0042] The weighted arbitration unit is configured to set priority weights for redundant data sources, with satellite remote sensing data and manual survey data having higher weights than radar estimated data. When the redundant data conflicts with the conclusions of the main monitoring data, a conflict evidence synthesis process based on DS evidence theory is triggered to reassign basic confidence levels. If the conflict cannot be resolved, it is marked as pending verification and a field review instruction is forcibly triggered.

[0043] Compared with the prior art, the present invention provides a real-time hydrological monitoring system, which has the following beneficial effects:

[0044] 1. In this invention, when monitoring watershed hydrological data in real time, a multi-source sensor adaptation unit uniformly receives heterogeneous hydrological data streams, and a spatiotemporal alignment unit performs timestamp synchronization and spatial coordinate mapping. This enables real-time verification and elimination of implicit data deviations caused by equipment differences and environmental interference, ensuring the accuracy of formatted hydrological data entering the system and preventing the accumulation of errors in subsequent hydrological topology modeling and confidence quantification, thereby improving the accuracy of hydrological anomaly detection.

[0045] 2. In this invention, when constructing the hydrological analysis model, key hydrological monitoring entities are identified through the monitoring node extraction unit, and edge sets are generated based on hydrological physical constraints through the dynamic adjacency calculation unit. The graph structure can be dynamically maintained according to the actual river confluence characteristics and spatial correlations, enabling the system to perceive and adapt to changes in watershed hydrological connections in real time. This solves the problem of the disconnect between static topology and actual physical processes. Furthermore, when adjacency relationships are abnormal, they can be corrected in real time through the graph structure update unit, ensuring the authenticity and reliability of the dynamic hydrological spatiotemporal adjacency graph.

[0046] 3. In this invention, when identifying and handling hydrological anomalies, suspected hydrological anomalies are extracted through a topology template matching unit and a confidence threshold screening unit. Cross-validation arbitration is performed by a redundant data source calling unit and a hydrological rule reasoning unit, enabling the system to integrate multi-source information and domain knowledge to make weighted decisions on anomalies. This solves the risk of misjudgment from a single information source and can accurately quantify flood and drought risks and match response strategies based on the final verification conclusions, further improving the reliability and scientific nature of early warning decisions. Attached Figure Description

[0047] Figure 1 This is a schematic diagram of the architecture of a real-time hydrological monitoring system according to the present invention;

[0048] Figure 2 This is a schematic diagram of the disassembled structure of the hydrological data access module of the present invention;

[0049] Figure 3 This is a schematic diagram of the split structure of the hydrological topology modeling module of the present invention. Detailed Implementation

[0050] 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.

[0051] Please see Figures 1-3 The specific implementation of a real-time hydrological monitoring system is as follows, including:

[0052] The hydrological data access module uses a multi-source sensor adaptation unit to receive heterogeneous hydrological data streams, performs timestamp synchronization and spatial coordinate mapping through a spatiotemporal alignment unit, and outputs formatted hydrological data through a data standardization unit.

[0053] The hydrological topology modeling module receives formatted hydrological data, identifies key hydrological monitoring entities through the monitoring node extraction unit, generates edge sets based on hydrological physical constraints through the dynamic adjacency calculation unit, and outputs a dynamic hydrological spatiotemporal adjacency graph through the graph structure update unit.

[0054] The confidence quantification module receives a dynamic hydrological spatiotemporal adjacency graph, calculates the spatial correlation of node data through the neighborhood consistency analysis unit, compares the hydrological evolution pattern using the hydrological prior verification unit, and outputs the joint confidence distribution of nodes and edges through the Bayesian fusion unit.

[0055] The anomaly pattern recognition module receives joint confidence distribution and formatted hydrological data, identifies the preset hydrological anomaly map structure through the topology template matching unit, extracts low-confidence node clusters through the confidence threshold screening unit, and outputs suspected hydrological anomaly events and propagation paths through the spatiotemporal evolution tracking unit.

[0056] The cross-validation arbitration module receives suspected hydrological anomalies, obtains independent monitoring data through the redundant data source calling unit, performs domain logic judgment using the hydrological rule reasoning unit, and generates the final verification conclusion and anomaly confidence level through the weighted arbitration unit.

[0057] The early warning decision support module receives the final verification conclusion, calculates the potential impact range through the flood and drought risk quantification unit, generates scheduling suggestions by querying the disposal strategy library through the contingency plan matching unit, and feeds back to the front-end hydrological monitoring platform through the instruction distribution unit.

[0058] The process in the hydrological data access module, which utilizes a multi-source sensor adaptation unit to receive heterogeneous hydrological data streams, performs timestamp synchronization and spatial coordinate mapping through a spatiotemporal alignment unit, and outputs formatted hydrological data through a data standardization unit, is as follows:

[0059] The multi-source sensor adapter unit is configured to be compatible with heterogeneous data protocols of rain gauges, water level gauges, flow meters, water quality sensors and mobile monitoring buoys, and to uniformly decode pulse, analog and digital signals;

[0060] The spatiotemporal alignment unit is configured to align asynchronous hydrological data points collected by different sensors to a standard time slice based on a unified hydrological time base and time window, and to assign unique geographic coordinates and watershed logical partition identifiers to each data source according to the preset watershed digital elevation model and hydrological station network topology.

[0061] The data standardization unit configuration is used to convert aligned hydrological data into a standardized data structure that includes physical dimensions, monitoring element types, spatial location codes, and timestamps, thereby eliminating dimensional differences.

[0062] The process of receiving formatted hydrological data in the hydrological topology modeling module, identifying key hydrological monitoring entities through the monitoring node extraction unit, generating edge sets based on hydrological physical constraints through the dynamic adjacency calculation unit, and outputting a dynamic hydrological spatiotemporal adjacency graph through the graph structure update unit is as follows:

[0063] The monitoring node extraction unit is configured to parse hydrological monitoring stations, reservoir gates, dangerous sections of dikes, and hydrological characteristic sections from formatted hydrological data as graph nodes, and extract the real-time hydrological element values ​​of the nodes;

[0064] The dynamic adjacency calculation unit is configured to define hydrological spatial adjacency rules and hydrological temporal adjacency rules. Spatial adjacency rules determine the physical association between nodes based on watershed connectivity, river flow direction, and preset distance thresholds. Temporal adjacency rules determine the temporal data association between nodes based on the propagation lag characteristics of hydrological process lines, and calculate the comprehensive adjacency strength by fusing spatial distance attenuation factors and temporal lag factors. The specific calculation process is as follows:

[0065] First, calculate the two nodes. and Spatial adjacency strength between The formula is:

[0066] ;

[0067] in, For nodes With nodes The actual distance traveled along the main channel of the river This is the spatial attenuation coefficient, used to characterize the attenuation effect of hydrological elements with spatial distance;

[0068] Secondly, calculate the temporal adjacency strength. The formula is:

[0069] ;

[0070] in, For nodes With nodes Monitor the difference in the time when the peak occurs. The theoretical confluence time is calculated based on the river's confluence velocity. This is the time decay coefficient;

[0071] Finally, the combined adjacency strength, which integrates the spatial distance attenuation factor and the time lag factor, is calculated. The formula is:

[0072] ;

[0073] in, This is a weighting coefficient, with a value range of [0, 1]. It is dynamically adjusted based on the current rainfall intensity in the watershed. When the rainfall intensity is high... Approaching 0 emphasizes time correlation; during periods of stable rainfall Approaching 1 to emphasize spatial relevance;

[0074] The graph structure update unit is configured to update when the overall adjacency strength is... When the connection threshold is exceeded, a weighted hydrological spatiotemporal adjacency edge is established between the two nodes, with the weight value proportional to the threshold. It also dynamically maintains the graph structure and updates weights as new data is input.

[0075] The confidence quantification module receives a dynamic hydrological spatiotemporal adjacency graph, calculates the spatial correlation of node data through a neighborhood consistency analysis unit, compares hydrological evolution patterns using a hydrological prior verification unit, and outputs the joint confidence distribution of nodes and edges via a Bayesian fusion unit.

[0076] The neighborhood consistency analysis unit is configured to target hydrological nodes. The hydrological data of all upstream and downstream adjacent nodes connected by hydrological spatiotemporal adjacency edges within the same time window are extracted. A multi-factor similarity algorithm is then used to calculate the similarity between the target node data and the data of each adjacent node. Specifically:

[0077] For time series data of water level and flow rate, the inverse of the dynamic time-normalized distance is used as a similarity measure, and the formula is:

[0078] ;

[0079] in, and Representing the target node respectively With neighboring nodes The sequence of hydrological elements within a time window It is a dynamic time-warped distance operator used to measure the morphological similarity between two time series curves;

[0080] After aggregation and normalization, an internal consistency score is generated, representing the consistency between the node's data and its hydrological neighborhood. The formula is:

[0081] ;

[0082] in, For the target node Internal consistency score, For nodes The set of adjacent nodes, For the weight of the connecting edges, This represents the number of adjacent nodes. For nodes With nodes The similarity between them;

[0083] The hydrological prior verification unit is configured to compare the current hydrological data of the target node with the theoretical range calculated based on historical hydrological characteristics, equipment calibration error range, and water balance physical equations, and output a prior conformity score. ;

[0084] The similarity between the target node data and the data of each neighboring node is calculated. After aggregation and normalization, an internal consistency score is generated that represents the consistency between the node data and its hydrological neighborhood.

[0085] The hydrological prior verification unit is configured to compare the current hydrological data of the target node with the theoretical value range calculated based on historical hydrological characteristic values, equipment calibration error range and water balance physical equation, and output the prior conformity score.

[0086] The Bayesian fusion unit is configured to construct a Bayesian network with node hydrological states as latent variables, and to incorporate internal consistency scores. Prior conformity score and data quality indicators As input for observational evidence, the joint probability distribution of nodes being in normal, suspicious, and abnormal states is inferred. And based on this, the confidence level of the hydrological correlation represented by the spatiotemporal adjacency edge of the hydrology is derived;

[0087] Among them, the Bayesian network adopts the Naive Bayes structure, that is, based on the hydrological state As the root node, with , , As leaf nodes, the conditional probability distribution between the root node and leaf nodes was obtained through offline training using historical hydrological datasets. Specifically, it involved statistically analyzing the distributions of conditional probability between historical normal and abnormal samples. , , The numerical distribution range is determined, and the conditional probability table is initialized using the maximum likelihood estimation method. During system operation, online fine-tuning is performed based on the data distribution within the sliding time window to ensure the timeliness of probability inference.

[0088] The anomaly pattern recognition module receives joint confidence distribution and formatted hydrological data, identifies the preset hydrological anomaly map structure through a topology template matching unit, extracts low-confidence node clusters using a confidence threshold filtering unit, and outputs suspected hydrological anomaly events and propagation paths through a spatiotemporal evolution tracking unit. The process is as follows:

[0089] The topology template matching unit is configured with a predefined hydrological anomaly map topology template containing flash floods, pollutant plume diffusion, waterlogging circles, and dam break chain transmission. It performs subgraph isomorphic matching between the dynamic hydrological spatiotemporal adjacency graph and the template to identify candidate anomaly clusters that conform to the anomaly spatial distribution characteristics.

[0090] Among them, the predefined hydrological anomaly map topology template is defined as a set of feature vectors. Each vector contains three quantitative indicators: node degree centrality, mean edge weight, and clustering coefficient. The VF2 algorithm is used for subgraph isomorphic matching to calculate the Euclidean distance between the topological feature vector of the dynamic hydrological spatiotemporal adjacency graph and the template feature vector. When the distance is less than the preset threshold, it is determined that there is a hydrological anomaly pattern in the region.

[0091] The confidence threshold filtering unit is configured to set a lower limit for node confidence and an upper limit for edge confidence, and to perform two-layer filtering on candidate abnormal clusters, removing nodes with low self-confidence and high confidence in their associations, and retaining low-confidence nodes with abnormal characteristics and questionable data reliability.

[0092] The spatiotemporal evolution tracking unit is configured to perform spatiotemporal clustering on the selected low-confidence nodes, trace the source of the anomaly and predict the scope of impact based on the direction and speed of hydrological propagation, and generate suspected hydrological anomaly events that include core anomaly nodes, affected watershed zoning and evolution trends.

[0093] The process by which the cross-validation arbitration module receives suspected hydrological anomalies, obtains independent monitoring data through the redundant data source calling unit, performs domain logic judgment using the hydrological rule reasoning unit, and generates the final validation conclusion and anomaly confidence level via the weighted arbitration unit is as follows:

[0094] The redundant data source calling unit is configured to parse the watershed range and element type involved in suspected hydrological anomalies, and retrieve independent third-party monitoring data for comparison from the preset redundant data source registry containing satellite remote sensing inversion data, meteorological radar rainfall data and manual patrol reporting data;

[0095] The hydrological rule reasoning unit is equipped with an expert rule base containing empirical formulas for rainstorm and flood forecasting, water level-discharge relationship curves, and flood control and drought relief dispatch procedures. It matches the characteristic parameters of suspected hydrological anomalies with the conditions in the rule base, performs logical deduction, and outputs verification opinions and weights based on domain knowledge.

[0096] The weighted arbitration unit is configured to comprehensively consider the anomaly degree of the main monitoring data, the consistency results of redundant data, and the expert rule reasoning opinions. It eliminates misjudgments from a single source of information through a weighted voting mechanism and outputs a final verification conclusion with a clear confidence level.

[0097] The process in the early warning decision support module—receiving the final verification conclusion, calculating the potential impact range through the flood and drought risk quantification unit, generating dispatch suggestions by querying the response strategy library using the contingency plan matching unit, and feeding back the suggestions to the front-end hydrological monitoring platform through the instruction distribution unit—is as follows:

[0098] The flood and drought risk quantification unit is configured to combine the underlying surface conditions of the watershed, the distribution of socio-economic assets, and the abnormal confidence level in the final verification conclusion to calculate the inundation range, water depth, and duration of flood and drought disasters, and to assess the risk level.

[0099] The contingency plan matching unit is configured to use the anomaly type, risk level, and protected objects within the scope of impact as indexes to query the flood control contingency plan database and match the corresponding engineering dispatch instructions, mass evacuation routes, and material allocation plans.

[0100] The instruction distribution unit is configured to encapsulate the final verification conclusions, risk quantification assessment reports, and disposal recommendations into standardized decision instruction packages, which are then distributed to monitoring and early warning platforms, reservoir dispatch centers, and emergency command terminals in relevant regions.

[0101] The process by which the dynamic adjacency computing unit generates edge sets based on hydrophysical constraints is further described as follows:

[0102] When configuring the dynamic adjacency calculation unit to calculate the comprehensive adjacency strength between any two hydrological monitoring nodes, the river confluence time parameter and energy attenuation coefficient are introduced, and the river energy attenuation characteristics are considered when calculating the spatial adjacency strength. Specifically:

[0103] In the formula Based on this, the energy attenuation correction factor derived from the Manning formula is introduced, and the formula for calculating spatial adjacency strength is optimized as follows:

[0104] ;

[0105] in, For nodes With nodes The actual distance traveled along the main channel of the river This is the roughness coefficient of the river channel. For hydraulic radius, The correction factor is used to reflect the energy dissipation characteristics of water flow as it propagates in the river channel, making the calculation of spatial adjacency intensity more consistent with hydraulic principles and improving the physical reality of the dynamic hydrological spatiotemporal adjacency diagram.

[0106] The spatial adjacency strength value is calculated by a negative exponential function of the actual flow distance between nodes along the main channel of the river. The temporal adjacency strength value is calculated by a negative exponential function of the difference between the peak occurrence times of monitoring between nodes and the deviation of the confluence time. The comprehensive adjacency strength is a weighted sum of the spatial adjacency strength and the temporal adjacency strength, and the weighting coefficient is dynamically adjusted with the rainfall intensity of the watershed.

[0107] The process of calculating the spatial correlation of node data in the neighborhood consistency analysis unit is further as follows:

[0108] The neighborhood consistency analysis unit is configured to automatically switch similarity algorithms based on the hydrological element type when calculating the data similarity between a target node and its neighboring nodes. For spatial distribution data of rainfall, a similarity algorithm based on Kriging interpolation residuals is used, specifically:

[0109] First, with the target node Centered on, using semi-mutation functions The spatial autocorrelation of fitted rainfall is given by the following formula:

[0110] ;

[0111] in, The value of the semi-variogram. The sampling interval distance. The interval is Number of sampling points For position Rainfall observation values ​​at the location, For position Rainfall observations at the location;

[0112] Then, the similarity of the Kriging interpolation residuals is calculated. The formula is:

[0113] ;

[0114] in, For spatial residual similarity based on Kriging interpolation results, For the target node The measured rainfall To utilize neighboring node data to node Kriging interpolation estimate of location, To estimate the standard deviation using interpolation;

[0115] For water level and flow time series, the inverse of dynamic time-normalized distance is used; for the spatial distribution of rainfall, Kriging interpolation residuals are used; and for water quality concentration, the inverse of relative deviation is used. Multiple sets of similarity values ​​are then weighted and averaged according to the importance of the watershed location of adjacent nodes to generate an internal consistency score. .

[0116] The process by which the weighted arbitration unit generates the final verification conclusion is further as follows:

[0117] The weighted arbitration unit is configured to set priority weights for redundant data sources, with satellite remote sensing data and manual survey data having higher weights than radar estimation data. When the conclusions of redundant data conflict with those of the main monitoring data, a conflict evidence synthesis process based on DS evidence theory is triggered. The specific synthesis rules are as follows:

[0118] Define the identification framework ,in This represents a normal state. This represents an abnormal state;

[0119] Let the basic confidence assignment function of the main monitoring data be: Redundant data sources are Then the basic confidence level assignment after synthesis The calculation formula is:

[0120] ;

[0121] in, Assign a base confidence level to the synthesized product. and Let represent the basic confidence assignment functions for the two sets of evidence from the primary monitoring data and the redundant data source, respectively. and Identify a subset of frame Θ. This usually represents an anomalous proposition. Represents a normal proposition. The conflict coefficient is calculated using the following formula: is the conflict coefficient, used to measure the degree of conflict between two sets of evidence;

[0122] If K is greater than the preset conflict threshold, the conflict is determined to be unresolved, marked as pending verification, and a forced on-site verification instruction is triggered. Otherwise, a final verification conclusion with a clear confidence level is output based on the synthesized confidence distribution.

[0123] The basic confidence level is reassigned. If the conflict cannot be resolved, it is marked as pending verification and an on-site verification instruction is forcibly triggered.

[0124] The operation steps of a real-time hydrological monitoring system are as follows:

[0125] Step 1: Standardization and preprocessing of heterogeneous hydrological data:

[0126] After the system starts up, it first performs data acquisition tasks through the hydrological data access module. The multi-source sensor adaptation unit is compatible with the heterogeneous communication protocols of rain gauges, water level gauges, current meters, water quality sensors and mobile monitoring buoys deployed in the watershed. It performs unified decoding of the acquired pulse, analog and digital signals. Subsequently, the spatiotemporal alignment unit aligns the asynchronous hydrological data points uploaded by different sensors to the standard time slice based on a unified hydrological time reference and time window. According to the preset watershed digital elevation model and hydrological station network topology, it assigns unique geographic coordinates and watershed logical partition identifiers to each data source. Finally, the data standardization unit converts the aligned hydrological data into formatted hydrological data containing physical dimensions, monitoring element types, spatial location codes and timestamps, eliminating dimensional differences and providing a standardized data foundation for subsequent analysis.

[0127] Step 2: Construction and maintenance of dynamic hydrological spatiotemporal adjacency graph:

[0128] The hydrological topology modeling module receives formatted hydrological data and begins constructing a hydrophysical model of the watershed. The monitoring node extraction unit extracts hydrological monitoring stations, reservoir gates, dangerous sections of dikes, and hydrological characteristic sections from the data as graph nodes, and extracts real-time hydrological element values ​​of the nodes. The dynamic adjacency calculation unit generates edge sets based on hydrophysical constraints. This unit defines hydrological spatial adjacency rules and hydrological temporal adjacency rules. Based on the watershed's river system connectivity, river flow direction, and the propagation lag characteristics of hydrological process lines, it calculates the comprehensive adjacency strength by fusing spatial distance attenuation factors and time lag factors. Finally, the graph structure update unit determines whether the comprehensive adjacency strength exceeds a preset connection threshold. If it does, a weighted hydrological spatiotemporal adjacency edge is established between the two nodes. The weight value is proportional to the comprehensive adjacency strength. The graph structure is dynamically maintained by adding, deleting, and updating weights as new data is continuously input, and a dynamic hydrological spatiotemporal adjacency graph that reflects the watershed status in real time is output.

[0129] Step 3: Quantitative evaluation of confidence based on neighborhood consistency:

[0130] The confidence quantification module receives a dynamic hydrological spatiotemporal adjacency graph and performs quality assessment on the node data. The neighborhood consistency analysis unit extracts the hydrological data of all upstream and downstream adjacent nodes connected by hydrological spatiotemporal adjacency edges within the same time window for the target hydrological node, calculates the similarity between the target node data and the data of each adjacent node, and generates an internal consistency score representing the consistency between the node data and its hydrological neighborhood after aggregation and normalization. The hydrological prior verification unit compares the current hydrological data of the target node with the theoretical value range calculated based on historical hydrological feature values, equipment calibration error range, and water balance physical equations, and outputs the prior conformity score. Finally, the Bayesian fusion unit constructs a Bayesian network with the node hydrological state as the latent variable, and uses the above internal consistency score, prior conformity score, and data quality index as input observation evidence to infer the joint probability distribution of the node being in a normal, suspicious, and abnormal state, and outputs the joint confidence distribution of the node and the edge.

[0131] Step 4: Identification and Spatiotemporal Evolution Tracking of Hydrological Anomaly Patterns:

[0132] The anomaly pattern recognition module combines joint confidence distribution with formatted hydrological data for anomaly detection. The topology template matching unit calls a predefined hydrological anomaly graph topology template that includes flash floods, pollutant plume diffusion, waterlogging circles, and dam-break chain transmission. It performs subgraph isomorphic matching between the dynamic hydrological spatiotemporal adjacency graph and the template to identify candidate anomaly clusters that conform to the anomaly spatial distribution characteristics. The confidence threshold screening unit sets a lower limit for node confidence and an upper limit for edge confidence, and performs double-layer filtering on the candidate anomaly clusters, removing nodes with too low self-confidence and too high confidence in their associations, and retaining low-confidence node clusters with anomaly characteristics and questionable data reliability. Subsequently, the spatiotemporal evolution tracking unit performs spatiotemporal clustering on the selected low-confidence nodes, traces the source of the anomaly and predicts the scope of impact based on the direction and speed of hydrological propagation, and generates suspected hydrological anomaly events and propagation paths that include core anomaly nodes, affected watershed zoning, and evolution trends.

[0133] Step 5: Cross-validation arbitration of multi-source information fusion:

[0134] The cross-validation arbitration module receives suspected hydrological anomalies to determine their authenticity. The redundant data source retrieval unit parses the watershed scope and element type involved in the event, and retrieves independent third-party monitoring data, including satellite remote sensing inversion data, meteorological radar rainfall data, and manual patrol reporting data, from the preset redundant data source registry for comparison. The hydrological rule reasoning unit uses an expert rule base containing empirical formulas for rainstorm and flood forecasting, water level-discharge relationship curves, and flood control and drought relief scheduling procedures to match the characteristic parameters of suspected events with the rule base conditions, perform logical deduction, and output verification opinions and weights based on domain knowledge. Finally, the weighted arbitration unit integrates the anomaly degree of the main monitoring data, the consistency results of redundant data, and the expert rule reasoning opinions, and eliminates misjudgments from a single information source through a weighted voting mechanism to generate a final verification conclusion and anomaly confidence level with a clear confidence level.

[0135] Step Six: Risk Quantification and Early Warning Decision-Making Instruction Distribution:

[0136] The early warning decision support module receives the final verification conclusion and initiates the emergency response process. The flood and drought risk quantification unit, combining the underlying surface conditions of the watershed, the distribution of socio-economic assets, and the anomaly confidence level in the final verification conclusion, calculates the inundation range, water depth, and duration of the flood and drought disaster, and assesses the risk level. The plan matching unit uses the anomaly type, risk level, and protected objects within the affected area as indexes to query the flood control plan database and match the corresponding engineering dispatch instructions, mass evacuation routes, and material allocation plans. Finally, the instruction distribution unit encapsulates the final verification conclusion, risk quantification assessment report, and disposal recommendations into a standardized decision instruction package and distributes it to the front-end hydrological monitoring platform, reservoir dispatch center, and emergency command terminal in the relevant areas, completing the closed-loop control from monitoring to decision-making.

[0137] It should be noted that, in this document, relational terms such as "first" and "second" are used merely to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0138] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A real-time hydrological monitoring system, characterized in that, include: The hydrological data access module uses a multi-source sensor adaptation unit to receive heterogeneous hydrological data streams, performs timestamp synchronization and spatial coordinate mapping through a spatiotemporal alignment unit, and outputs formatted hydrological data through a data standardization unit. The hydrological topology modeling module receives the formatted hydrological data, identifies key hydrological monitoring entities through the monitoring node extraction unit, generates edge sets based on hydrological physical constraints through the dynamic adjacency calculation unit, and outputs a dynamic hydrological spatiotemporal adjacency graph through the graph structure update unit. The confidence quantification module receives the dynamic hydrological spatiotemporal adjacency graph, calculates the spatial correlation of node data through the neighborhood consistency analysis unit, compares the hydrological evolution pattern through the hydrological prior verification unit, and outputs the joint confidence distribution of nodes and edges through the Bayesian fusion unit. The anomaly pattern recognition module receives the joint confidence distribution and formatted hydrological data, identifies the preset hydrological anomaly map structure through the topology template matching unit, extracts low-confidence node clusters through the confidence threshold filtering unit, and outputs suspected hydrological anomaly events and propagation paths through the spatiotemporal evolution tracking unit. The cross-validation arbitration module receives the suspected hydrological anomaly event, obtains independent monitoring data through the redundant data source calling unit, performs domain logic judgment using the hydrological rule reasoning unit, and generates the final verification conclusion and anomaly confidence level through the weighted arbitration unit. The early warning decision support module receives the final verification conclusion, calculates the potential impact range through the flood and drought risk quantification unit, generates scheduling suggestions by querying the disposal strategy library through the contingency plan matching unit, and feeds them back to the front-end hydrological monitoring platform through the instruction distribution unit.

2. The real-time hydrological monitoring system according to claim 1, characterized in that, The process by which the hydrological data access module receives heterogeneous hydrological data streams using a multi-source sensor adaptation unit, performs timestamp synchronization and spatial coordinate mapping through a spatiotemporal alignment unit, and outputs formatted hydrological data through a data standardization unit is as follows: The multi-source sensor adapter unit is configured to be compatible with heterogeneous data protocols of rain gauges, water level gauges, flow meters, water quality sensors and mobile monitoring buoys, and to uniformly decode pulse, analog and digital signals; The spatiotemporal alignment unit is configured to align asynchronous hydrological data points collected by different sensors to a standard time slice based on a unified hydrological time base and time window, and to assign unique geographic coordinates and watershed logical partition identifiers to each data source according to the preset watershed digital elevation model and hydrological station network topology. The data standardization unit configuration is used to convert aligned hydrological data into a standardized data structure that includes physical dimensions, monitoring element types, spatial location codes, and timestamps.

3. The real-time hydrological monitoring system according to claim 1, characterized in that, The process by which the hydrological topology modeling module receives the formatted hydrological data, identifies key hydrological monitoring entities through the monitoring node extraction unit, generates edge sets based on hydrological physical constraints through the dynamic adjacency calculation unit, and outputs a dynamic hydrological spatiotemporal adjacency graph through the graph structure update unit is as follows: The monitoring node extraction unit is configured to parse hydrological monitoring stations, reservoir gates, dangerous sections of dikes, and hydrological characteristic sections from formatted hydrological data as graph nodes, and extract the real-time hydrological element values ​​of the nodes; The dynamic adjacency calculation unit is configured to define hydrological spatial adjacency rules and hydrological temporal adjacency rules. The spatial adjacency rules determine the physical correlation between nodes based on the connectivity of the watershed system, the direction of the river channel, and a preset distance threshold. The temporal adjacency rules determine the temporal correlation of data between nodes based on the propagation lag characteristics of the hydrological process line, and calculate the comprehensive adjacency strength by fusing the spatial distance attenuation factor and the temporal lag factor. The graph structure update unit is configured to establish a weighted hydrological spatiotemporal adjacency edge between two nodes when the overall adjacency strength exceeds a preset connection threshold. The weight value is proportional to the overall adjacency strength, and the addition, deletion, and weight update of the graph structure are dynamically maintained as new data is input.

4. The real-time hydrological monitoring system according to claim 1, characterized in that, The confidence quantification module receives the dynamic hydrological spatiotemporal adjacency graph, calculates the spatial correlation of node data through the neighborhood consistency analysis unit, compares the hydrological evolution patterns using the hydrological prior verification unit, and outputs the joint confidence distribution of nodes and edges through the Bayesian fusion unit. The neighborhood consistency analysis unit is configured to extract the hydrological data of all upstream and downstream adjacent nodes connected by hydrological spatiotemporal adjacency edges within the same time window for the target hydrological node, calculate the similarity between the target node data and the data of each adjacent node, and generate an internal consistency score that represents the consistency between the node data and its hydrological neighborhood after aggregation and normalization. The hydrological prior verification unit is configured to compare the current hydrological data of the target node with the theoretical value range calculated based on historical hydrological characteristic values, equipment calibration error range and water balance physical equation, and output the prior conformity score. The Bayesian fusion unit is configured to construct a Bayesian network with the hydrological state of the nodes as the latent variable. The internal consistency score, prior conformity score and data quality index are used as the input of observation evidence to infer the joint probability distribution of the nodes in normal, suspicious and abnormal states, and the confidence of the hydrological association represented by the hydrological spatiotemporal adjacency edge is derived from this.

5. A real-time hydrological monitoring system according to claim 1, characterized in that, The process by which the anomaly pattern recognition module receives the joint confidence distribution and formatted hydrological data, identifies the preset hydrological anomaly map structure through the topology template matching unit, extracts low-confidence node clusters using the confidence threshold filtering unit, and outputs suspected hydrological anomaly events and propagation paths through the spatiotemporal evolution tracking unit is as follows: The topology template matching unit is configured with a predefined hydrological anomaly map topology template that includes flash floods, pollutant plume diffusion, waterlogging circles, and dam break chain transmission. The dynamic hydrological spatiotemporal adjacency graph is matched with the template to identify candidate anomaly clusters that conform to the characteristics of anomaly spatial distribution. The confidence threshold filtering unit is configured to set a lower limit for node confidence and an upper limit for edge confidence, and to perform two-layer filtering on candidate abnormal clusters, removing nodes with low self-confidence and high confidence in their associations, and retaining low-confidence nodes with abnormal characteristics and questionable data reliability. The spatiotemporal evolution tracking unit is configured to perform spatiotemporal clustering on the selected low-confidence nodes, trace the source of the anomaly and predict the scope of impact based on the direction and speed of hydrological propagation, and generate suspected hydrological anomaly events that include core anomaly nodes, affected watershed zoning and evolution trends.

6. A real-time hydrological monitoring system according to claim 1, characterized in that, The process by which the cross-validation arbitration module receives the suspected hydrological anomaly event, obtains independent monitoring data through the redundant data source calling unit, performs domain logic judgment using the hydrological rule reasoning unit, and generates the final verification conclusion and anomaly confidence level through the weighted arbitration unit is as follows: The redundant data source calling unit is configured to parse the watershed range and element type involved in suspected hydrological anomalies, and retrieve independent third-party monitoring data for comparison from the preset redundant data source registry containing satellite remote sensing inversion data, meteorological radar rainfall data and manual patrol reporting data; The hydrological rule reasoning unit is equipped with an expert rule base containing empirical formulas for rainstorm and flood forecasting, water level-discharge relationship curves, and flood control and drought relief dispatch procedures. It matches the characteristic parameters of suspected hydrological anomalies with the conditions in the rule base, performs logical deduction, and outputs verification opinions and weights based on domain knowledge. The weighted arbitration unit is configured to comprehensively consider the anomaly degree of the main monitoring data, the consistency results of redundant data, and the expert rule reasoning opinions. It eliminates misjudgments from a single source of information through a weighted voting mechanism and outputs a final verification conclusion with a clear confidence level.

7. A real-time hydrological monitoring system according to claim 1, characterized in that, The process by which the early warning decision support module receives the final verification conclusion, calculates the potential impact range through the flood and drought risk quantification unit, generates scheduling suggestions by querying the disposal strategy library using the contingency plan matching unit, and feeds the suggestions back to the front-end hydrological monitoring platform through the instruction distribution unit is as follows: The flood and drought risk quantification unit is configured to combine the underlying surface conditions of the watershed, the distribution of socio-economic assets, and the abnormal confidence level in the final verification conclusion to calculate the inundation range, water depth, and duration of flood and drought disasters, and to assess the risk level. The contingency plan matching unit is configured to use the anomaly type, risk level, and protected objects within the scope of impact as indexes to query the flood control contingency plan database and match the corresponding engineering dispatch instructions, mass evacuation routes, and material allocation plans. The instruction distribution unit is configured to encapsulate the final verification conclusions, risk quantification assessment reports, and disposal recommendations into standardized decision instruction packages, which are then distributed to monitoring and early warning platforms, reservoir dispatch centers, and emergency command terminals in relevant regions.

8. A real-time hydrological monitoring system according to claim 3, characterized in that, The process by which the dynamic adjacency calculation unit generates edge sets based on hydrophysical constraints is further as follows: The dynamic adjacency calculation unit is configured to incorporate river confluence time parameters and energy attenuation coefficients when calculating the comprehensive adjacency strength between any two hydrological monitoring nodes. The spatial adjacency strength value is calculated as a negative exponential function of the actual flow distance between nodes along the main channel. The temporal adjacency strength value is calculated as a negative exponential function of the difference between the peak monitoring times between nodes and the deviation of the confluence time. The comprehensive adjacency strength is a weighted sum of spatial and temporal adjacency strengths, and the weighting coefficient is dynamically adjusted with the rainfall intensity in the basin.

9. A real-time hydrological monitoring system according to claim 4, characterized in that, The process by which the neighborhood consistency analysis unit calculates the spatial correlation of node data is further as follows: The neighborhood consistency analysis unit is configured to automatically switch similarity algorithms according to the type of hydrological element when calculating the data similarity between the target node and its neighboring nodes. For water level and flow time series, dynamic time-normalized distance reciprocal is used; for spatial distribution of rainfall, Kriging interpolation residual is used; and for water quality concentration, relative deviation reciprocal is used. The obtained multiple sets of similarity values ​​are then weighted and averaged according to the importance of the watershed location of the neighboring nodes to generate an internal consistency score.

10. A real-time hydrological monitoring system according to claim 6, characterized in that, The process by which the weighted arbitration unit generates the final verification conclusion is further as follows: The weighted arbitration unit is configured to set priority weights for redundant data sources, with satellite remote sensing data and manual survey data having higher weights than radar estimated data. When the redundant data conflicts with the conclusions of the main monitoring data, a conflict evidence synthesis process based on DS evidence theory is triggered to reassign basic confidence levels. If the conflict cannot be resolved, it is marked as pending verification and a field review instruction is forcibly triggered.

Images