Intelligent supervision method and system based on data visualization platform

Abstract

The invention relates to the technical field of data processing, in particular to an intelligent supervision method and system based on a data visualization platform, and the method comprises the steps: collecting multi-source water conservancy data through dual-channel redundancy check, converting the multi-source water conservancy data into a standardized space-time matrix, carrying out the data cleaning through a containerization adaptation and variational auto-encoder, and marking the confidence; hierarchically storing the data based on a confidence threshold, and generating a routing strategy through reinforcement learning to distribute data fragments; calculating a flood peak evolution prediction result by using space-time diagram convolution, calculating an output risk assessment value by combining a Bayesian network and long and short-term memory, and activating federal learning parameter updating when the confidence coefficient is insufficient; the digital twin model loads physical constraint parameters to deduce a flood control scene, and a scheduling instruction set and a rehearsal animation are generated through multi-objective optimization; and dynamically rendering to generate an interactive visual interface, associating a causal deduction path, and triggering federal model parameter updating by user feedback, thereby realizing dynamic collaboration of the forecasting and early warning rehearsal plan function chain.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, and in particular to an intelligent monitoring method and system based on a data visualization platform. Background Technology

[0002] The water resources data visualization platform integrates multiple water resources monitoring data sources, clearly displays the dynamic changes of hydrological elements through a graphical interface, and continuously analyzes real-time data streams through the integration of machine learning models in the intelligent monitoring system. It automatically detects abnormal flow or rainfall deviations, and then predicts potential disaster risks such as floods or droughts. This capability helps water resource managers make rapid response decisions, optimize reservoir scheduling or water-saving measures, and ultimately improve water resource management efficiency and reduce the probability of disaster impact.

[0003] The lack of modular decoupling design and standardized data conversion mechanisms in existing water conservancy data visualization systems leads to the failure of intelligent linkage of all elements of watershed data. Specifically, when the reservoir scheduling module urgently needs to access meteorological rainfall data, river water level data, and soil moisture data for flood risk simulation, the existing architecture forces manual reconstruction of the data pipeline for adaptation. At this time, satellite remote sensing imagery data with inconsistent structures cannot be converted into standardized raster data that can be resolved by hydrological models in real time. Simultaneously, the downstream disaster loss assessment module remains blocked due to its reliance on upstream preprocessing results. For example, in a sudden heavy rainfall event in a certain watershed, format conflicts caused by diverse data sources (such as misaligned timestamps between drone aerial images and ground sensor data) force the system to interrupt the "forecast-simulation" business chain, directly delaying the flood storage and dispatch decision-making window and causing an overall decline in disaster prevention and control effectiveness. Summary of the Invention

[0004] To address the shortcomings of existing technologies, this invention provides an intelligent monitoring method and system based on a data visualization platform. This method solves the technical problems caused by the strong correlation between functional modules hindering independent expansion, the difficulty of real-time synchronization of multi-dimensional water conservancy data, and the blocking effect caused by the linear dependence of the analysis process, which leads to the dynamic collaborative failure of the functional chain of basin-level forecasting, early warning, simulation, and contingency plan.

[0005] To solve the above-mentioned technical problems, the specific details of the present invention are as follows: In a first aspect, the present invention provides an intelligent supervision method based on a data visualization platform, comprising: Step 1: Collect meteorological radar data, river water level sensor data, soil moisture monitoring data and satellite remote sensing images, and output the initial dataset with time alignment using dual-channel redundancy verification; Step 2: Convert the initial dataset into a standardized spatiotemporal matrix, including: containerization to adapt to a unified data format, pre-training a variational autoencoder to calculate reconstruction error, adversarial generation to scan outlier data points, and outputting a cleaned spatiotemporal matrix with confidence labels; Step 3: Receive the cleaned spatiotemporal matrix, store the matrix data in layers according to the confidence threshold, store data with a confidence level greater than or equal to the threshold in the high-speed cache library, store historical data in the spatiotemporal index library, monitor the real-time business scenario status and data storage location, generate a routing strategy table through reinforcement learning decision, and extract data from the storage library and distribute it to the message queue according to the routing strategy table. Step 4: Extract data shards from the message queue and input them into parallel processing. Parallel processing includes: spatiotemporal graph convolution calculation to generate flood peak evolution prediction results, Bayesian network and long short-term memory calculation to output risk assessment values, activation of federated learning parameter updates when the ensemble voting confidence is lower than the threshold, and output of analysis results with reliability indicators. Step 5: Receive the analysis results output in Step 4, load the pre-stored physical constraint parameters into the digital twin model, deduce the flood control scenario based on the analysis results, output the dynamic parameters of flood peak evolution, input the dynamic parameters of flood peak evolution into multi-objective optimization calculation, iteratively generate the scheduling instruction set, and generate the pre-show animation data based on the scheduling instruction set; Step 6: Dynamically render and load the scheduling instruction set and pre-rendering animation data, generate an interactive visualization interface and associate it with the causal inference path, and receive user feedback instructions to trigger the update of federated model parameters.

[0006] Furthermore, in the intelligent supervision method based on a data visualization platform described in this invention, step 1 includes: The main channel transmits the raw data stream to the platform entry point, while the auxiliary channel simultaneously transmits the feature hash value to the platform entry point. The platform entry point receives main channel data and calculates the real-time hash value of the main channel data. Extract the feature hash value transmitted through the secondary channel, compare the real-time hash value with the feature hash value. If they are inconsistent, trigger the data source node to retransmit the main channel data until the hash values ​​are consistent, and output the initial dataset with integrity markers.

[0007] Furthermore, in the intelligent supervision method based on a data visualization platform described in this invention, step 2 includes: Receive the initial dataset, containerize and adapt the initial dataset, parse the meteorological data to obtain the spatiotemporal matrix of rainfall intensity, perform band calculations on the remote sensing data to obtain the elevation raster, and integrate the spatiotemporal matrix of rainfall intensity and the elevation raster into a unified format matrix. The format-unified matrix is ​​input into the variational autoencoder to calculate the reconstruction error. The format-unified matrix is ​​input into the generative adversarial network to scan abnormal data points and output anomaly scores. The reconstruction error and anomaly scores are merged to generate data confidence labels. Data confidence labels are mapped to a format-unified matrix, and the cleaned spatiotemporal matrix with confidence labels is output.

[0008] Furthermore, in the intelligent supervision method based on a data visualization platform described in this invention, step 3 includes: After receiving the cleaned spatiotemporal matrix, the data is classified and stored according to the confidence threshold. Data with a confidence level greater than or equal to the threshold is stored in the high-speed cache and a spatiotemporal index is created. The remaining data is stored in the historical database. Based on the stored data location, read the location status, capture the real-time business scenario code and data volume, and input the location status, business scenario code and data volume into the reinforcement learning decision. The reward function calculates the optimal routing action, generates a routing policy table, retrieves data shards from the repository based on the routing policy table, and distributes the data shards to the message queue.

[0009] Furthermore, in the intelligent supervision method based on a data visualization platform described in this invention, step 4 includes: Receive the data fragments distributed in step 3, and construct a watershed topology graph based on the water conservancy facility relationships in the data fragments. Nodes are water conservancy facility entities, and edges are water flow directions and distances. The watershed topology map is input into the spatiotemporal graph for convolution calculation, and the flood peak evolution prediction results are output. Input the flood peak evolution prediction results into the risk assessment and output the risk assessment value; When the confidence level of the integrated voting for the risk assessment value is lower than the threshold, the federated learning parameter update is activated. The system integrates flood peak evolution prediction results, risk assessment values, and federated learning update signals to output analysis results with reliability indicators.

[0010] Furthermore, in the intelligent supervision method based on a data visualization platform described in this invention, step 5 includes: Receive the analysis results output from step 4, load the pre-stored physical constraint parameters into the digital twin model, deduce the flood control scenario based on the analysis results, and output the dynamic parameters of the flood peak evolution; The dynamic parameters of the flood peak evolution are input into multi-objective optimization calculations, and chromosome encoding is used to generate a gate opening sequence. The fitness calculation simultaneously evaluates flood control safety and water supply security, iteratively optimizes the gate opening sequence, outputs the Pareto optimal scheduling scheme, and generates pre-show animation data based on the Pareto optimal scheduling scheme.

[0011] Furthermore, in the intelligent supervision method based on a data visualization platform described in this invention, step 5 includes: Receive the initial gate opening sequence, and use tournament selection to screen parent chromosomes; The parent chromosome is input into a simulated binary crossover to generate the offspring chromosome sequence; The weighting coefficients α, β, and γ of the offspring chromosome sequence are dynamically adjusted, the fitness calculation rules are updated, the offspring chromosome sequence is iteratively optimized based on the update rules, and Pareto optimal scheduling data is output. Personnel transfer path data is generated based on Pareto optimal scheduling data.

[0012] Furthermore, in the intelligent supervision method based on a data visualization platform described in this invention, step 6 includes: Dynamically render and load the scheduling instruction set and pre-show animation data to generate an interactive visualization interface, and overlay a flood peak risk heat map on the digital elevation model base map; The interactive visual interface receives user click commands on high-risk areas, triggering a replay of the digital twin model's simulation process; The visualization interface displays the planned path animation, which shows the iterative trajectory of the multi-objective optimization algorithm. The interactive visualization interface receives user correction commands, generates a difference vector, encrypts and uploads the difference vector, and triggers the update of the federated model parameters.

[0013] Furthermore, in the intelligent supervision method based on a data visualization platform described in this invention, step 6 further includes: Receive encrypted uploaded difference vectors, perform federated aggregation and parsing of the difference vectors, and calculate the global model parameter update amount; Distribute global model parameter updates to independent processing; Independently process and update the parameters of the risk assessment model to generate updated risk assessment values; The updated risk assessment values ​​are input into the dynamic rendering, and the risk heatmap on the visualization interface is refreshed in real time.

[0014] Secondly, the present invention provides an intelligent monitoring system based on a data visualization platform, applied to the intelligent monitoring method based on a data visualization platform as described above, comprising: The data acquisition unit is configured to acquire meteorological radar data, river water level sensor data, soil moisture monitoring data and satellite remote sensing images, and perform dual-channel redundancy check to output time-aligned initial datasets. The data processing unit is configured to convert the initial dataset into a standardized spatiotemporal matrix, including: containerizing and adapting to a unified data format, calculating the reconstruction error using a pre-trained variational autoencoder, generating adversarial data to scan outlier data points, and outputting a cleaned spatiotemporal matrix with confidence labels. The data analysis unit is configured to receive the cleaned spatiotemporal matrix, store the matrix data in layers according to the confidence threshold, monitor the real-time business scenario status and data storage location, generate a routing strategy table through reinforcement learning decision-making, and extract data from the repository and distribute it to the message queue according to the routing strategy table. The parallel processing unit is configured to extract data shards from the message queue, perform spatiotemporal graph convolution calculations to generate flood peak evolution prediction results, calculate risk assessment values ​​using Bayesian networks and long short-term memory, activate federated learning parameter updates when the ensemble voting confidence is below a threshold, and output analysis results with reliability indicators. The generation unit is configured to receive analysis results, load pre-stored physical constraint parameters into the digital twin model to simulate flood control scenarios, output dynamic parameters of flood peak evolution, perform multi-objective optimization calculations to iteratively generate a scheduling instruction set, and generate pre-simulation animation data based on the scheduling instruction set. The optimization unit is configured to dynamically render and load the scheduling instruction set and pre-rendering animation data, generate an interactive visualization interface and associate it with the causal inference path, and receive user feedback instructions to trigger the update of federated model parameters.

[0015] Beneficial effects of this invention; This invention effectively eliminates the risk of packet loss in multi-source water conservancy data transmission through a dual-channel redundancy verification mechanism. A containerized adaptation process combined with intelligent quality assessment resolves the conflict between satellite imagery and sensor data formats. The routing strategy table generated by reinforcement learning decision-making supports hot-swapping of microservice modules, thus eliminating strong inter-module dependencies. In the parallel processing flow, spatiotemporal graph convolution calculation and Bayesian risk assessment are executed synchronously to avoid linear blocking in the analysis process. Federated learning parameter updates automatically activate when the confidence level of the integrated voting is insufficient, improving model reliability. The digital twin model loads physical constraint parameters to deduce flood control scenarios and combines multi-objective optimization calculations to output Pareto optimal scheduling schemes, achieving millisecond-level dynamic coordination of the forecasting, early warning, rehearsal, and contingency plan functional chains. User feedback-driven federated model parameter updates and real-time updates of the visualization interface form a closed loop for decision optimization, fundamentally solving the problem of failure in the coordination of the four pre-flood control functional chains at the basin level, and significantly improving the effectiveness of flood disaster prevention and control. Attached Figure Description

[0016] To more clearly illustrate the technical solution of the present invention, the drawings used in the embodiments will be briefly introduced below. Obviously, for those skilled in the art, other drawings can be obtained based on the drawings without creative effort.

[0017] Figure 1 This is a flowchart illustrating an intelligent supervision method based on a data visualization platform, provided as an embodiment of the present invention. Detailed Implementation

[0018] To make the technical solution of the present invention clearer, the present invention will be clearly and completely described below with reference to specific embodiments and corresponding drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them. 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. The present invention provided by various embodiments will be described in detail below with reference to the accompanying drawings. To better understand the purpose of the present invention, the present invention will be described in further detail below.

[0019] Firstly, please refer to Figure 1 This invention provides an intelligent supervision method based on a data visualization platform, comprising: Step 1: Collect meteorological radar data, river water level sensor data, soil moisture monitoring data and satellite remote sensing images, and output the initial dataset with time alignment using dual-channel redundancy verification; Step 2: Convert the initial dataset into a standardized spatiotemporal matrix, including: containerization to adapt to a unified data format, pre-training a variational autoencoder to calculate reconstruction error, adversarial generation to scan outlier data points, and outputting a cleaned spatiotemporal matrix with confidence labels; Step 3: Receive the cleaned spatiotemporal matrix, store the matrix data in layers according to the confidence threshold, store data with a confidence level greater than or equal to the threshold in the high-speed cache library, store historical data in the spatiotemporal index library, monitor the real-time business scenario status and data storage location, generate a routing strategy table through reinforcement learning decision, and extract data from the storage library and distribute it to the message queue according to the routing strategy table. Step 4: Extract data shards from the message queue and input them into parallel processing. Parallel processing includes: spatiotemporal graph convolution calculation to generate flood peak evolution prediction results, Bayesian network and long short-term memory calculation to output risk assessment values, activation of federated learning parameter updates when the ensemble voting confidence is lower than the threshold, and output of analysis results with reliability indicators. Step 5: Receive the analysis results output in Step 4, load the pre-stored physical constraint parameters into the digital twin model, deduce the flood control scenario based on the analysis results, output the dynamic parameters of flood peak evolution, input the dynamic parameters of flood peak evolution into multi-objective optimization calculation, iteratively generate the scheduling instruction set, and generate the pre-show animation data based on the scheduling instruction set; Step 6: Dynamically render and load the scheduling instruction set and pre-rendering animation data, generate an interactive visualization interface and associate it with the causal inference path, and receive user feedback instructions to trigger the update of federated model parameters.

[0020] During the data acquisition phase, the system synchronously acquires meteorological radar data, river water level sensor data, soil moisture monitoring data, and satellite remote sensing imagery through a distributed sensor network. A dual-channel redundancy verification mechanism employs a parallel transmission method with primary and secondary channels. The primary channel transmits the raw data stream, while the secondary channel synchronously transmits feature hash values. Upon receiving the data, the platform's input terminal calculates the real-time hash value and compares it with the hash value from the secondary channel. If they do not match, the data source is retransmitted, thus ensuring data integrity and time alignment. This design solves the synchronization failure problem caused by packet loss or timing misalignment in multi-source water conservancy data transmission, providing reliable input for subsequent processing.

[0021] The initial dataset enters the data transformation stage, where a containerization adaptation process unifies the format of heterogeneous data sources. Meteorological data is parsed into a spatiotemporal matrix of rainfall intensity, and remote sensing data is generated into an elevation raster through band operations. A variational autoencoder calculates the reconstruction error to quantify data quality, and an adversarial generative network scans for sensor anomalies to output anomaly scores. The two are combined to generate data confidence labels. The cleaned spatiotemporal matrix with confidence labels is output, providing a data quality basis for hierarchical storage.

[0022] After cleaning, the spatiotemporal matrix is ​​stored using a tiered strategy based on confidence thresholds. High-confidence data is stored in a high-speed cache and a spatiotemporal index is created, while historical data is archived in the spatiotemporal index repository. The system monitors the business scenario status and data storage location in real time, and the reinforcement learning decision module generates a routing strategy table based on scenario encoding, data volume, and storage status. Data shards are extracted from the corresponding repository according to the strategy table and distributed to the message queue, achieving dynamic scheduling of data resources and module decoupling.

[0023] Data sharding in the message queue triggers parallel processing. Spatiotemporal graph convolution computation, based on the watershed topology, weighted aggregates the features of adjacent nodes to capture the spatiotemporal patterns of flood peak evolution and output prediction results. Bayesian networks quantify uncertainty factors, and long short-term memory networks learn temporal dependencies to jointly generate risk assessment values. When the ensemble voting confidence level falls below a threshold, the federated learning parameter update mechanism is activated, improving model reliability through multi-node collaborative training, and finally outputting analysis results with reliability indicators.

[0024] The analysis results drive the digital twin model to simulate flood control scenarios. The model loads pre-stored physical constraint parameters, such as hydraulic equations and topographic data, and combines them with dynamic parameters of flood peak evolution to simulate the evolution of flood control scenarios. Multi-objective optimization calculations use chromosome encoding to generate gate opening sequences, and fitness functions simultaneously evaluate flood control safety and water supply security. Through iterative optimization, the Pareto optimal scheduling scheme is output, and pre-simulation animation data is generated to intuitively demonstrate the scheduling effect.

[0025] The dynamic rendering module loads the scheduling instruction set and pre-simulation animation data, overlaying a flood peak risk heat map onto the digital elevation model base map to generate an interactive visualization interface. Users can click on high-risk areas through the interface to trigger a replay of the digital twin simulation process, viewing the animation of the contingency plan path and the iteration trajectory of the optimization algorithm; user correction instructions generate difference vectors, which are encrypted and uploaded to trigger the update of federated model parameters, forming a closed-loop collaborative mechanism from decision-making simulation to feedback optimization.

[0026] Specifically, in the intelligent supervision method based on a data visualization platform described in this invention, step 1 includes: The main channel transmits the raw data stream to the platform entry point, while the auxiliary channel simultaneously transmits the feature hash value to the platform entry point. The platform entry point receives main channel data and calculates the real-time hash value of the main channel data. Extract the feature hash value transmitted through the secondary channel, compare the real-time hash value with the feature hash value. If they are inconsistent, trigger the data source node to retransmit the main channel data until the hash values ​​are consistent, and output the initial dataset with integrity markers.

[0027] During the data acquisition phase, a dual-channel redundancy verification mechanism transmits the raw data stream to the platform's entry point via the main channel, while the auxiliary channel transmits feature hash values ​​in parallel. The main channel carries complete data packets of meteorological radar data, river water level sensing data, soil moisture monitoring data, and satellite remote sensing imagery, while the auxiliary channel transmits lightweight feature summaries generated based on the SHA-256 algorithm. This dual-path parallel transmission architecture effectively avoids the risk of data packet loss due to network jitter or equipment failure in single-channel transmission.

[0028] Upon receiving data from the main channel, the platform's entry point immediately calculates the real-time hash value. The calculation process uses the same hash algorithm as the data source to ensure consistency in the comparison benchmark. After the real-time hash value is generated, the system automatically extracts the feature hash values ​​pre-transmitted from the auxiliary channel for comparison. The hash comparison module employs a direct memory comparison method, avoiding latency caused by disk read / write operations and meeting the real-time requirements of water conservancy monitoring scenarios.

[0029] When the real-time hash value and the characteristic hash value are inconsistent, the system determines that the data transmission integrity has been compromised and immediately triggers the data source node retransmission mechanism. The retransmission command is sent to the edge data acquisition node through a dedicated control channel, requesting the retransmission of the main channel data stream. During the retransmission process, hash value comparison is continuously performed until the two hash values ​​are completely consistent. Finally, an initial dataset with an integrity tag is output, which includes the timestamp of hash verification success and data volume information, providing a reliable basis for subsequent processing.

[0030] This verification mechanism is particularly suitable for cross-regional water conservancy monitoring networks. For example, when rain gauges in mountainous areas transmit data via wireless networks, signal interference may cause data packet loss. The dual-channel design allows the system to quickly detect anomalies by comparing the hash values ​​of the secondary channel even if some data is lost on the primary channel, and trigger the retransmission of the specified data segment, thus avoiding the bandwidth waste caused by existing end-to-end retransmissions.

[0031] Specifically, in the intelligent supervision method based on a data visualization platform described in this invention, step 2 includes: Receive the initial dataset, containerize and adapt the initial dataset, parse the meteorological data to obtain the spatiotemporal matrix of rainfall intensity, perform band calculations on the remote sensing data to obtain the elevation raster, and integrate the spatiotemporal matrix of rainfall intensity and the elevation raster into a unified format matrix. The format-unified matrix is ​​input into the variational autoencoder to calculate the reconstruction error. The format-unified matrix is ​​input into the generative adversarial network to scan abnormal data points and output anomaly scores. The reconstruction error and anomaly scores are merged to generate data confidence labels. Data confidence labels are mapped to a format-unified matrix, and the cleaned spatiotemporal matrix with confidence labels is output.

[0032] In the data transformation phase, the system initiates a containerized adaptation and transformation process after receiving the initial dataset. This process deploys dedicated parsing containers for heterogeneous data sources such as meteorological radar data and river water level sensor data. The meteorological data container parses GRIB format files to generate a spatiotemporal matrix of rainfall intensity, while the remote sensing data container extracts elevation raster information through multispectral band operations. The outputs of each container adhere to a unified coordinate system and timestamp specification, ultimately integrating into a unified matrix with consistent dimensions. This design solves the integration challenge caused by format differences between satellite imagery and sensor data. For example, when there are spatial scale inconsistencies between UAV aerial photography data and ground monitoring data, containerized adaptation automatically performs coordinate transformation and data resampling to ensure matrix element alignment.

[0033] The format-unified matrix is ​​then input in parallel into two quality assessment channels. The variational autoencoder calculates the reconstruction error based on a pre-trained model, quantifying the degree of data distortion by comparing the input data with the decoded output. The generative adversarial network (GAN) identifies anomalous patterns through a discriminator, outputting anomaly scores for each data point. The two assessment results are weighted and merged in the confidence fusion module. The reconstruction error reflects the level of systematic error, while the anomaly scores detect sudden interference, together forming the basis for generating data confidence labels. For example, in a water conservancy monitoring scenario, the variational autoencoder can mark outliers in the rainfall intensity matrix that exceed historical patterns, while the GAN can identify abrupt changes in the elevation grid.

[0034] The generated confidence labels are associated with a standardized matrix via a spatial mapping module. Each grid cell is appended with a corresponding confidence value, forming a labeled spatiotemporal matrix data structure. This design allows downstream processing modules to flexibly adjust data processing strategies based on confidence thresholds, such as initiating a review mechanism for low-confidence regions or allocating more computing resources for verification. The output cleaned spatiotemporal matrix with confidence labels not only includes standardized monitoring data but also carries complete quality assessment information, providing structured input for subsequent hierarchical storage and parallel processing.

[0035] Specifically, in the intelligent supervision method based on a data visualization platform described in this invention, step 3 includes: After receiving the cleaned spatiotemporal matrix, the data is classified and stored according to the confidence threshold. Data with a confidence level greater than or equal to the threshold is stored in the high-speed cache and a spatiotemporal index is created. The remaining data is stored in the historical database. Based on the stored data location, read the location status, capture the real-time business scenario code and data volume, and input the location status, business scenario code and data volume into the reinforcement learning decision. The reward function calculates the optimal routing action, generates a routing policy table, retrieves data shards from the repository based on the routing policy table, and distributes the data shards to the message queue.

[0036] In the data storage stage, after receiving the cleaned spatiotemporal matrix, the system implements a hierarchical storage strategy based on a preset confidence threshold. High-confidence data is stored in a high-speed cache library with a spatiotemporal index, while historical data is stored in a spatiotemporal index library. This design optimizes resource allocation based on data quality. For example, when real-time rainfall data is needed for flood peak prediction, the system prioritizes retrieving high-confidence data from the high-speed cache library to reduce access latency.

[0037] The system continuously monitors the real-time status of business scenarios and data storage locations, capturing key parameters such as business scenario codes and data volume. The monitoring module acquires business scenario codes via sensor networks, such as typhoon warnings or drought assessments, while simultaneously reading the status of data storage locations, including the load on cache and index libraries. Parameters are input into the reinforcement learning decision-making module, which calculates the optimal routing action based on the Q-learning algorithm and generates a routing strategy table.

[0038] The reinforcement learning decision-making module evaluates the utility of different routing actions through a reward function, such as routing high-priority data to parallel processing flows. The routing policy table explicitly specifies which repository to retrieve data shards from and how to distribute them to the message queue. Based on the policy table, the system retrieves data shards from the corresponding repository and distributes them to downstream processing modules via the message queue, achieving dynamic scheduling of the data flow.

[0039] In water conservancy monitoring applications, when the system detects a flood warning scenario, reinforcement learning decision-making prioritizes routing real-time water level data in segments to the flood peak evolution prediction module, ensuring that critical data is processed in a timely manner. This dynamic routing mechanism effectively avoids data congestion and improves system response speed.

[0040] Specifically, in the intelligent supervision method based on a data visualization platform described in this invention, step 4 includes: Receive the data fragments distributed in step 3, and construct a watershed topology graph based on the water conservancy facility relationships in the data fragments. Nodes are water conservancy facility entities, and edges are water flow directions and distances. The watershed topology map is input into the spatiotemporal graph for convolution calculation, and the flood peak evolution prediction results are output. Input the flood peak evolution prediction results into the risk assessment and output the risk assessment value; When the confidence level of the integrated voting for the risk assessment value is lower than the threshold, the federated learning parameter update is activated. The system integrates flood peak evolution prediction results, risk assessment values, and federated learning update signals to output analysis results with reliability indicators.

[0041] Receive the data fragments distributed in step 3, parse the spatial association attributes of water conservancy facilities in the data fragments, construct a watershed topology map with reservoirs and sluice gates as node entities and river flow direction and hydraulic connection distance as edge relationships; The watershed topology map is input into the spatiotemporal graph convolutional computation module. The graph convolutional layer aggregates the hydrological features of adjacent nodes, and the temporal convolutional layer captures the temporal pattern of flood peak propagation, outputting the flood peak evolution prediction results. The flood peak evolution prediction results are input into the risk assessment module. The Bayesian network calculates the probability distribution of disaster occurrence, the long short-term memory network learns the historical flood sequence pattern, and the risk assessment value is output by fusion. When the integrated voting confidence of the risk assessment value is lower than the preset threshold, the federated learning parameter update mechanism is activated to coordinate multiple local nodes to update the encrypted model parameters. The system integrates flood peak evolution prediction results, risk assessment values, and federated learning parameter update signals, adds timestamps and credibility markers, and outputs analysis results with reliability indicators.

[0042] Specifically, in the intelligent supervision method based on a data visualization platform described in this invention, step 5 includes: Receive the analysis results output from step 4, load the pre-stored physical constraint parameters into the digital twin model, deduce the flood control scenario based on the analysis results, and output the dynamic parameters of the flood peak evolution; The dynamic parameters of the flood peak evolution are input into multi-objective optimization calculations, and chromosome encoding is used to generate a gate opening sequence. The fitness calculation simultaneously evaluates flood control safety and water supply security, iteratively optimizes the gate opening sequence, outputs the Pareto optimal scheduling scheme, and generates pre-show animation data based on the Pareto optimal scheduling scheme.

[0043] In the flood control decision-making stage, after receiving the analysis results output from step 4, the digital twin model loads pre-stored physical constraint parameters. These parameters include fundamental physical rules such as hydraulic equation coefficients, topographic elevation data, and river roughness coefficients. Based on the flood peak evolution prediction and risk assessment values ​​in the analysis results, the model simulates the evolution of flood control scenarios in a virtual environment. The simulation process modeles the reservoir storage and release dynamics and river flood propagation under different scheduling strategies, outputting dynamic parameters of flood peak evolution, including key indicators such as water level change rate and flood peak propagation time.

[0044] The multi-objective optimization calculation module for inputting dynamic parameters of flood peak evolution uses chromosome encoding technology to generate gate opening sequences. Each chromosome represents a set of gate scheduling schemes, with gene loci corresponding to gate opening values ​​for a specific time period. The fitness calculation module simultaneously evaluates two objective functions: flood control safety and water supply security. Flood control safety measures the risk of river water levels exceeding warning levels, while water supply security assesses the reservoir's ability to meet demand.

[0045] The gate opening sequence is iteratively optimized through selection, crossover, and mutation operations. A tournament selection mechanism selects parent individuals with high fitness, a binary crossover operation is simulated to generate the offspring population, and a dynamic mutation operation maintains population diversity. Individual fitness is recalculated in each iteration, gradually approaching the Pareto optimal solution set. When the maximum number of iterations or the convergence threshold is reached, the Pareto optimal scheduling scheme set is output.

[0046] Preliminary animation data is generated based on the Pareto optimal scheduling scheme. The animation sequence showcases the flood evolution process under different scheduling strategies, including visualization elements such as reservoir water level change curves, river flood peak propagation trajectories, and dynamic changes in inundation areas. The animation data is linked to the scheduling instruction set, providing basic materials for subsequent visualization.

[0047] In specific application scenarios, when the system handles watershed rainstorm events, the digital twin model extrapolates the flood evolution process over the next 72 hours based on real-time rainfall forecasts. The multi-objective optimization module simultaneously considers downstream flood control safety and urban water supply needs, generating the optimal gate control sequence that balances flood discharge and water storage. The final output simulation animation can intuitively demonstrate the implementation effects of different decision-making schemes, supporting flood control consultation and decision-making.

[0048] Specifically, in the intelligent supervision method based on a data visualization platform described in this invention, step 5 includes: Receive the initial gate opening sequence, and use tournament selection to screen parent chromosomes; The parent chromosome is input into a simulated binary crossover to generate the offspring chromosome sequence; The weighting coefficients α, β, and γ of the offspring chromosome sequence are dynamically adjusted, the fitness calculation rules are updated, the offspring chromosome sequence is iteratively optimized based on the update rules, and Pareto optimal scheduling data is output. Personnel transfer path data is generated based on Pareto optimal scheduling data.

[0049] In the optimization calculation phase, the system initiates the genetic algorithm optimization process after receiving the initial gate opening sequence. A tournament selection mechanism selects parent chromosomes with high fitness from the population; each chromosome represents a set of gate scheduling schemes. The selection process simulates a competition mechanism, randomly selecting multiple individuals to compare fitness values. The winning individual serves as the parent for reproduction, maintaining population diversity and avoiding premature convergence.

[0050] The parent chromosome input simulates a binary crossover operation, generating offspring chromosome sequences through gene position exchange. The crossover operation randomly selects crossover points based on a probability threshold, exchanging parent chromosome segments to produce new individuals. The simulated binary crossover uses continuous variable processing to keep gene values ​​within a reasonable range, avoiding illegal solutions and enhancing the algorithm's global search capability.

[0051] The weighting coefficients α, β, and γ of the offspring chromosome sequences are dynamically adjusted, with each coefficient corresponding to a weighting proportion in objective functions such as flood control safety and water supply security. The weighting adjustment module adaptively updates the coefficient values ​​based on the iteration progress and population distribution, balancing conflicting relationships in multi-objective optimization. The updated weighting coefficients are then input into the fitness calculation rules to reassess the overall fitness of each chromosome.

[0052] Based on the updated fitness calculation rules, the system iteratively optimizes the offspring chromosome sequence. Each iteration performs selection, crossover, and mutation operations, gradually approaching the Pareto optimal solution set. When the maximum number of iterations or the solution set convergence threshold is reached, a Pareto optimal scheduling dataset is output, including multiple non-dominated solution schemes.

[0053] Based on Pareto optimal scheduling data, the system generates personnel evacuation route data. The route planning module combines flood inundation range prediction results, transportation network data, and shelter location information to calculate the optimal evacuation route. The output data includes evacuation time windows, route capacity constraints, and emergency resource allocation plans, forming a complete personnel dispatch plan.

[0054] Specifically, in the intelligent supervision method based on a data visualization platform described in this invention, step 6 includes: Dynamically render and load the scheduling instruction set and pre-show animation data to generate an interactive visualization interface, and overlay a flood peak risk heat map on the digital elevation model base map; The interactive visual interface receives user click commands on high-risk areas, triggering a replay of the digital twin model's simulation process; The visualization interface displays the planned path animation, which shows the iterative trajectory of the multi-objective optimization algorithm. The interactive visualization interface receives user correction commands, generates a difference vector, encrypts and uploads the difference vector, and triggers the update of the federated model parameters.

[0055] In the visualization stage, the system's dynamic rendering module loads the scheduling instruction set and pre-simulation animation data, generating an interactive visualization interface through the WebGL graphics engine. The interface uses a digital elevation model as its base map, overlaid with a flood peak risk heat map. The heat map's color gradient reflects the distribution of risk assessment values, with red areas marking high-risk inundation zones. The rendering process integrates terrain elevation data and dynamic parameters of flood peak evolution in real time, generating a 3D scene visualization effect that allows users to zoom and rotate to view watershed details.

[0056] The interactive visual interface receives user commands to click on high-risk areas, and the interface event capture module parses the geographic location corresponding to the click coordinates. The click command triggers the digital twin model's simulation playback function. The playback module loads historical simulation data and recreates the entire flood peak evolution process in a timeline sliding format. The playback process simultaneously displays water level change curves and gate operation sequences to assist users in analysis and decision-making.

[0057] The visualization interface continuously plays animated sequences of planned routes, showcasing the iterative trajectory of a multi-objective optimization algorithm. The evolutionary process of the genetic algorithm is presented as a dynamic scatter plot, with each particle representing a scheduling scheme and the trajectory reflecting the Pareto front convergence path. Simultaneously, animations of personnel transfer routes are displayed, with arrows indicating evacuation directions and path color intensity representing passage priority.

[0058] The interactive visual interface provides a parameter correction panel, accepting correction commands from users by dragging sliders or inputting numerical values. Users can adjust parameters such as gate opening thresholds and evacuation time windows. The correction command processing module calculates the difference vector between the new and original schemes in real time. The difference vector includes parameter offsets and timing adjustments, which are encrypted using the AES encryption algorithm and then uploaded to the cloud server.

[0059] The encrypted upload of the difference vector triggers the federated model parameter update mechanism. The federated aggregation server receives encrypted data uploaded by each terminal, parses the difference vector through homomorphic decryption, and calculates the global model parameter update. The updated model parameters are distributed to each edge node, optimizing the inference rules of the risk assessment module in real time, forming a user feedback-driven model evolution closed loop.

[0060] Specifically, in the intelligent supervision method based on a data visualization platform described in this invention, step 6 further includes: Receive encrypted uploaded difference vectors, perform federated aggregation and parsing of the difference vectors, and calculate the global model parameter update amount; Distribute global model parameter updates to independent processing; Independently process and update the parameters of the risk assessment model to generate updated risk assessment values; The updated risk assessment values ​​are input into the dynamic rendering, and the risk heatmap on the visualization interface is refreshed in real time.

[0061] During the federated model update phase, the system receives encrypted difference vectors uploaded from multiple terminal nodes. These difference vectors are uploaded to the federated aggregation server via a secure transmission protocol, and each vector includes local adjustments to the model parameters made by the user. The encryption layer employs an asymmetric encryption algorithm to protect data transmission, preventing sensitive information from being intercepted or tampered with during the transmission link.

[0062] The federated aggregation server initiates the parsing process, performing batch decryption of the received encrypted difference vectors. The parsing module extracts the parameter offsets and adjustment weights from the vectors, and calculates the global model parameter update based on the federated averaging algorithm. The calculation process comprehensively considers the data quality weights of each terminal node to avoid excessive impact of individual abnormal data on the global model.

[0063] The calculated global model parameter updates are distributed to each independent processing node via a secure channel. The distribution mechanism employs an incremental update strategy, transmitting only the parameter changes rather than the complete model to reduce network bandwidth consumption. Each independent processing node receives the updates and merges them with its local model parameters to complete the model version iteration.

[0064] The independent processing node re-runs the risk assessment calculation based on the updated model parameters. The node loads the latest water monitoring data and generates updated risk assessment values ​​through a forward propagation algorithm. The calculation process integrates real-time hydrological characteristics and historical patterns, outputting a timestamped risk probability distribution.

[0065] The updated risk assessment values ​​are input into the dynamic rendering engine in real time. The rendering engine adjusts the color mapping of the heatmap based on the new risk assessment values, displaying high-risk areas in warm colors and low-risk areas in cool colors. The visualization interface refreshes the risk heatmap frame-by-frame, ensuring that users always see results based on the latest model assessment.

[0066] In watershed flood control applications, when multiple flood control command centers make different risk assessments for the same area, the system integrates the local experience of each center through a federated aggregation mechanism to generate a more accurate global risk assessment. The updated risk heat map can display the dynamic changes in risk level during the evolution of the flood peak in real time, providing a reference for continuous optimization of flood control decisions.

[0067] Secondly, the present invention provides an intelligent monitoring system based on a data visualization platform, applied to the intelligent monitoring method based on a data visualization platform as described above, comprising: The data acquisition unit is configured to acquire meteorological radar data, river water level sensor data, soil moisture monitoring data and satellite remote sensing images, and perform dual-channel redundancy check to output time-aligned initial datasets. The data processing unit is configured to convert the initial dataset into a standardized spatiotemporal matrix, including: containerizing and adapting to a unified data format, calculating the reconstruction error using a pre-trained variational autoencoder, generating adversarial data to scan outlier data points, and outputting a cleaned spatiotemporal matrix with confidence labels. The data analysis unit is configured to receive the cleaned spatiotemporal matrix, store the matrix data in layers according to the confidence threshold, monitor the real-time business scenario status and data storage location, generate a routing strategy table through reinforcement learning decision-making, and extract data from the repository and distribute it to the message queue according to the routing strategy table. The parallel processing unit is configured to extract data shards from the message queue, perform spatiotemporal graph convolution calculations to generate flood peak evolution prediction results, calculate risk assessment values ​​using Bayesian networks and long short-term memory, activate federated learning parameter updates when the ensemble voting confidence is below a threshold, and output analysis results with reliability indicators. The generation unit is configured to receive analysis results, load pre-stored physical constraint parameters into the digital twin model to simulate flood control scenarios, output dynamic parameters of flood peak evolution, perform multi-objective optimization calculations to iteratively generate a scheduling instruction set, and generate pre-simulation animation data based on the scheduling instruction set. The optimization unit is configured to dynamically render and load the scheduling instruction set and pre-rendering animation data, generate an interactive visualization interface and associate it with the causal inference path, and receive user feedback instructions to trigger the update of federated model parameters.

[0068] In its implementation, the system is deployed at the basin flood control command center, connecting to data sources such as meteorological radar stations, hydrological monitoring stations, and satellite receiving stations. A distributed architecture is adopted, with the main processing node deployed at the provincial center and edge nodes deployed at the municipal-level branch centers, forming a hierarchical processing network. The main node is equipped with a high-performance graphics server cluster, and the edge nodes are configured with real-time data processing modules, interconnected via a dedicated fiber optic network.

[0069] The data acquisition process integrates the base data stream from a meteorological radar, with a volumetric scan data set every 6 minutes. Simultaneously, it receives minute-level monitoring values ​​from a river water level sensor. The soil moisture monitor uploads a data set every half hour, and satellite remote sensing imagery updates elevation information every 3 hours. The main channel uses TCP to transmit raw data packets, while the auxiliary channel synchronously sends data feature summaries via UDP. A data verification module at the platform's entry point calculates SHA-256 hash values ​​in real time for integrity verification, immediately triggering retransmission at a designated node upon detecting data packet loss.

[0070] The data transformation phase deploys a containerized adaptation cluster, including independent modules such as a meteorological data parsing container, a remote sensing image processing container, and a sensor data transformation container. The meteorological data container parses radar-based data to generate a 1km×1km grid of rainfall intensity matrix, while the remote sensing container processes Sentinel satellite imagery to generate a 30-meter resolution elevation raster. The outputs of each container are uniformly transformed to the WGS84 coordinate system, with time aligned to the hour, forming a standard spatiotemporal matrix. The quality assessment module uses a pre-trained variational autoencoder to calculate data reconstruction errors, a generative adversarial network to detect sensor outliers, and assigns a confidence score to each data point.

[0071] The cleaned spatiotemporal matrix is ​​stored in a hierarchical storage system. Real-time data with a confidence level of 0.9 or higher is stored in a Redis cache, and a spatiotemporal composite index is established. Historical data is stored in a time-series database, with index keys including timestamps and spatial coordinates. The routing decision module monitors business scenarios in real time. When the system identifies a typhoon warning pattern, it automatically routes real-time rainfall data to the flood peak prediction pipeline; when a drought warning is issued, it prioritizes routing soil moisture data to the drought assessment module.

[0072] After the parallel processing unit extracts data shards from the message queue, it first constructs a topological map of the watershed's water conservancy facilities. Reservoirs are used as nodes, and river channels as edges. Node attributes include reservoir capacity curves, and edge attributes include river channel gradient and roughness. A spatiotemporal graph convolutional network employs a 3-layer graph convolutional structure combined with temporal convolution to aggregate upstream node features and predict downstream flood peak evolution. A Bayesian network calculates conditional probabilities, and a long short-term memory network learns historical flood sequences. When the confidence score of multiple models falls below 0.85, a federated learning mechanism is triggered to update the model parameters.

[0073] The digital twin model loads physical parameters such as river topography and reservoir characteristic curves to simulate flood control scenarios in a virtual environment. The multi-objective optimization module uses the NSGA-II algorithm with a population size of 100 and iterates 200 times to find the Pareto optimal solution. The fitness function considers both the downstream flood control safety index and the urban water supply guarantee rate, outputting the optimal gate control strategy that balances flood discharge and water storage.

[0074] The visualization interface uses WebGL technology to render a 3D scene, overlaying a flood peak risk heat map onto the digital elevation model. Users can click on high-risk areas to replay the digital twin simulation process and view the flood's trajectory. After receiving user feedback on gate opening adjustments, the system generates a parameter difference vector, updates the risk assessment model through a federated learning mechanism, and refreshes the heat map display in real time.

[0075] In responding to basin-wide rainstorm events, the system achieves a complete closed loop from data collection to decision optimization through the aforementioned process. For example, when radar detects the movement of a heavy rainfall belt, the system automatically triggers the flood peak prediction process, uses a digital twin model to predict the flood evolution over the next 6 hours, optimizes algorithms to generate gate scheduling schemes, and provides a visual interface to intuitively display the effects of different decisions, thus forming a collaborative system of forecasting, early warning, rehearsal, and contingency plans.

[0076] This invention systematically addresses the issue of functional chain coordination failure. At the module decoupling level, a containerized adaptation process and a message queue mechanism are employed to enable independent expansion of functional modules. Each data source type corresponds to an independent container instance; for example, a meteorological data parsing container and a remote sensing data processing container run in parallel, exchanging data between containers through standardized interfaces. The message queue acts as a data relay station, allowing asynchronous communication between the flood peak prediction module and the drought assessment module. New functional modules only need to subscribe to the corresponding message queue topic to access the system, eliminating compile-level dependencies between modules.

[0077] To address the challenge of synchronizing multi-source data, a dual-channel redundancy check and intelligent quality assessment mechanism is designed. The main channel transmits the raw data stream, while the auxiliary channel synchronously transmits lightweight feature hash values. The platform's entry point detects data packet loss through real-time hash comparison and immediately triggers retransmission. During the data conversion stage, a variational autoencoder is introduced to calculate reconstruction errors, an adversarial generative network is used to detect abnormal patterns, and a confidence label is assigned to each data point. When timestamps conflict between satellite imagery and sensor data, the system automatically aligns the time reference and fills in missing values, ensuring spatiotemporal consistency of multi-source data.

[0078] To prevent process bottlenecks, a dynamic collaborative architecture of parallel processing and federated verification is constructed. A spatiotemporal graph convolutional network processes watershed topology relationships to generate flood peak predictions, while Bayesian networks and long short-term memory networks compute risk assessment values ​​in parallel. Dual-path analysis eliminates serial dependencies. When the risk assessment confidence level falls below a threshold, the federated learning mechanism is automatically activated, with each edge node updating model parameters locally and coordinating with the server to aggregate gradients and update the global model, avoiding performance bottlenecks caused by centralized processing. The digital twin model loads physical constraint parameters to simulate multiple scenarios in real time, and multi-objective optimization algorithms iterate in parallel to find the optimal solution, forming a non-blocking pipeline from forecast to contingency plan.

[0079] Specifically, in the context of river basin flood control, when a typhoon warning is triggered, the system dynamically expands the resources of the flood peak prediction module through a containerized architecture, while maintaining the normal operation of the drought assessment module. A dual-channel verification mechanism ensures that meteorological radar data and water level monitoring data arrive synchronously, and the intelligent quality assessment module identifies abnormal water level values ​​and marks them as low-confidence. Parallel processing synchronously generates flood peak evolution predictions and inundation risk maps. When multiple model results diverge, a federated learning mechanism integrates local knowledge from various flood control command centers to quickly update the models, ensuring the output of reliable dispatch plans before the arrival of heavy rain. This design enables the forecasting, early warning, rehearsal, and contingency plan functional chains to form a dynamic collaborative closed loop, effectively improving emergency response efficiency.

[0080] Embodiment 1 of this invention: In a scenario of responding to a watershed-wide rainstorm event, the system is deployed at a river flood control command center, connecting meteorological radar stations, water level monitoring stations, soil moisture meters, and remote sensing satellite data sources. When the radar detects the movement of a heavy rainfall belt, the data acquisition unit initiates a dual-channel redundancy verification mechanism: the main channel transmits the radar base data stream, while the auxiliary channel synchronously sends data feature hash values. The platform entry point calculates the real-time hash value and compares it with the auxiliary channel summary. If a data packet is lost, it immediately triggers retransmission at the edge nodes, outputting a time-aligned initial dataset. The data processing unit parses the radar data through a containerized adaptation process to generate a 1-kilometer grid rainfall intensity matrix, processes satellite imagery to generate a 30-meter elevation raster, calculates the reconstruction error to mark low-quality data points using a variational autoencoder, identifies abnormal water level values ​​using a generative adversarial network, and outputs a standardized matrix with confidence labels. The data analysis unit stores real-time data with a confidence level of 0.9 or higher in a cache library, and the reinforcement learning module generates a routing strategy based on the typhoon warning scenario, distributing data fragments to the flood peak prediction message queue. The parallel processing unit constructs a watershed topology map, a spatiotemporal graph convolutional network aggregates upstream reservoir features to predict flood peak evolution, and a Bayesian network calculates the inundation risk probability. When the risk assessment confidence level falls below a threshold, a federated learning mechanism is activated to update model parameters. The generation unit loads river channel topography parameters to simulate flood control scenarios, and a multi-objective optimization algorithm iterates 200 times to generate gate scheduling schemes, outputting a pre-simulation animation to display the flood evolution trajectory. The optimization unit renders a 3D risk heat map using a WebGL engine. Users can click on high-risk areas to replay the simulation process, adjust gate opening parameters to generate difference vectors, encrypt and upload them, triggering federated model updates and real-time refresh of the visualization interface.

[0081] Embodiment 2 of this invention: In a regional drought monitoring scenario, the system is deployed at a water resource dispatch center and connected to a soil moisture monitoring network and meteorological forecast data. When satellite remote sensing indicates that soil moisture is below a threshold, the data acquisition unit simultaneously receives soil moisture meter data and meteorological forecasts, with dual-channel verification ensuring data integrity. The data processing unit parses multispectral remote sensing data into a soil moisture raster using independent containers, adapts the containerization to a unified elevation data and sensor data format, uses a variational autoencoder to identify abnormal drought spots, employs an adversarial generative network to detect sensor drift errors, and outputs a soil moisture matrix with confidence labels. The data analysis unit stores data hierarchically according to confidence levels, and the reinforcement learning module routes data to the drought assessment pipeline based on drought early warning scenarios. The parallel processing unit integrates historical drought patterns with real-time data, uses a long short-term memory network to predict drought development trends, a Bayesian network to calculate water shortage risks, and a federated learning mechanism to integrate experiences from multiple irrigation districts to update the assessment model. The generation unit loads reservoir storage curves and water demand parameters, uses a multi-objective optimization algorithm to generate water resource allocation plans, balances agricultural irrigation and urban water supply needs, and outputs an animation of drought mitigation plans. The optimization unit dynamically displays reservoir water level changes and water conveyance routes, allows users to adjust water supply ratios to generate parameter differences, and uses federated aggregation to update model parameters, optimizing drought risk assessment results in real time.

[0082] The technical terms used in this invention are explained below: A variational autoencoder (VAE) is a generative model that maps input data to a probability distribution in a latent space using an encoder, and then reconstructs the original data by sampling from this distribution using a decoder. In this invention, the VAE is pre-trained on a standardized spatiotemporal matrix, and the reconstruction error is calculated by comparing the difference between the original matrix and the decoder output matrix. This error value is used to quantify the reliability of each data point; a larger reconstruction error indicates a greater deviation of the data point from the normal distribution, which may include noise or outliers. The reconstruction error calculation result serves as one of the important indicators for data quality assessment, providing a basic input for the subsequent generation of data confidence labels.

[0083] Generative Adversarial Networks (GANs) consist of a generator and a discriminator. The generator attempts to produce samples similar to the real data distribution, while the discriminator strives to distinguish between real and generated samples. This invention utilizes the discriminator's discriminative power within the GAN as an anomaly detector to scan for anomalous data points in the spatiotemporal matrix. The discriminator outputs an anomaly score for each data point; a higher score indicates a greater likelihood of the point belonging to an anomalous pattern. This mechanism is particularly suitable for detecting anomalies caused by sensor malfunctions or sudden environmental changes, anomalies that may not be effectively identified using existing reconstruction error methods.

[0084] Reinforcement learning learns optimal strategies through interaction between an agent and its environment, aiming to maximize cumulative rewards. In the data routing stage of this invention, the reinforcement learning agent uses data storage location, real-time business scenario status, and data volume as state inputs, and the selection of routing targets for data shards as its action space. The agent evaluates the effectiveness of different routing actions through a reward function; for example, routing high-priority data to a parallel processing flow can yield a positive reward. Through continuous trial and error learning, the intelligent learning decision module generates an optimal routing strategy table, achieving dynamic intelligent scheduling of data resources.

[0085] Spatiotemporal graph convolution is a deep learning model specifically designed for processing graph-structured data, capable of simultaneously capturing spatial topological relationships and temporal dynamic changes. This invention abstracts watershed water conservancy facilities into a graph structure, where nodes represent facilities such as reservoirs, and edges represent the direction and distance of water flow. The spatiotemporal graph convolution model aggregates the neighborhood features of each node through graph convolutional layers to capture spatial dependencies; and analyzes the changing patterns of each node's features over time through temporal convolutional layers to capture temporal dependencies. This spatiotemporal joint modeling capability effectively predicts the evolution of flood peaks in the watershed, outputting the flood peak propagation trajectory and intensity changes over a future period.

[0086] Bayesian networks are probabilistic graphical models that represent conditional dependencies between variables using directed acyclic graphs (DAGs) and perform probabilistic inference. In the risk assessment phase of this invention, the Bayesian network uses meteorological factors, hydrological data, and topographic features as nodes to construct conditional probabilistic relationships between these factors and flood risk. When new observational data is input, the network updates the probability distribution of risk events through a probability propagation algorithm, outputting a probability-based risk assessment value. This probability-based assessment method can clearly quantify the uncertainty of predictions, providing more comprehensive reference information for decision-making.

[0087] Long Short-Term Memory (LSTM) networks are a special type of recurrent neural network that effectively solves the gradient vanishing problem in long sequence training through gating mechanisms. This invention utilizes LTM networks to learn long-term dependencies in historical disaster events, capturing the temporal patterns of flood disaster occurrence and development. By inputting continuous time-series hydrological and meteorological data, the network learns the complete cyclical pattern of flood peak formation, evolution, and dissipation. When receiving real-time data streams, the network can predict future risk development trends based on the learned temporal patterns, complementing Bayesian networks to jointly improve the accuracy of risk assessment.

[0088] Federated learning is a distributed machine learning paradigm that allows multiple participants to train models locally without exchanging raw data, achieving collaborative modeling only by exchanging model parameter updates. In this invention, the system activates the federated learning mechanism when the risk assessment results of multiple terminal nodes differ significantly or have insufficient confidence. Each edge node calculates the model parameter update on its local data, encrypts it, and uploads it to the coordination server for secure aggregation, generating an improved global model. This mechanism protects the data privacy of each node while improving the model's generalization ability and reliability through knowledge sharing.

[0089] Digital twin models integrate physical laws, geometric features, and behavioral rules to construct a digital mapping in virtual space that perfectly corresponds to a physical entity. The digital twin model in this invention loads physical constraint parameters such as river topography, reservoir characteristics, and hydraulic equations to construct a virtual replica of the watershed. When real-time monitoring data or forecast results are input, the model performs simulations based on physical laws, simulating the flood propagation process under different scenarios. This physical mechanism-based simulation method effectively compensates for the shortcomings of purely data-driven models, improving the interpretability and reliability of forecast results.

[0090] Multi-objective optimization algorithms aim to find a set of solutions that can simultaneously optimize multiple conflicting objectives, i.e., a Pareto optimal solution set. This invention employs a multi-objective optimization method based on a genetic algorithm, encoding the gate opening sequence as chromosomes, with each gene representing the gate opening at a specific time period. The optimization process simultaneously considers multiple objectives such as flood control safety, water supply security, and power generation efficiency, continuously evolving the population through operations such as selection, crossover, and mutation. Ultimately, a set of non-dominated solutions is output, forming multiple scheduling schemes for decision-makers to choose from, thereby maximizing overall benefits.

[0091] The construction of a digital twin model begins with a comprehensive digital mapping of the physical world. The model first loads pre-stored watershed physical constraint parameters, including river channel topographic elevation data, reservoir characteristic curves, hydraulic equation coefficients, soil permeability, and structural parameters of hydraulic engineering facilities—fundamental physical rules. These parameters are acquired through multiple channels, including geographic information systems, engineering design drawings, and field survey data, and undergo standardized processing to form a unified digital description. The model architecture adopts a hierarchical design: the bottom layer is a geospatial database, the middle layer is a physical law engine, and the top layer is a dynamic simulation core, forming a complete framework from static data to dynamic deduction.

[0092] In the data processing path, the digital twin model receives analysis results from the parallel processing unit, including flood peak evolution prediction data and risk assessment indicators. The model first performs data fusion processing, aligning real-time monitoring data with the prediction results spatiotemporally to eliminate scale differences and temporal biases between different data sources. During the fusion process, a Kalman filter algorithm is used to smooth the data, reducing the impact of measurement noise on the extrapolation results while maintaining sensitivity to sudden changes. The fused data generates an input dataset with a consistent spatiotemporal benchmark, providing a high-quality data foundation for subsequent scenario extrapolation.

[0093] After data fusion, the model initiates flood control scenario simulation calculations. The simulation process is based on fundamental principles of hydrology and hydraulics, employing the Saint-Venant equations to describe the movement of floodwaters in the river channel, and using the finite volume method for discrete solution. The model calculates hydraulic elements such as water level, flow rate, and velocity in each river segment in real time, simulating the propagation of flood waves in the basin. Simultaneously, the influence of engineering measures such as reservoir regulation, gate control, and dike constraints is considered, achieving dynamic interaction between engineering scheduling and flood evolution through coupled solution. The simulation time step is adaptively adjusted according to the urgency of the flood; a smaller step size is used in rapidly changing flow regions to ensure computational stability, while a larger step size is used in gently changing flow regions to improve computational efficiency.

[0094] During the simulation process, the model undergoes continuous real-time correction. By assimilating the latest monitoring data and comparing the differences between the simulation results and actual observations, the model parameters are dynamically adjusted to reduce simulation errors. The correction mechanism employs an ensemble Kalman filter method, continuously optimizing model state variables and key parameters through parallel simulations of multiple parameter sets and observation data assimilation. This real-time correction mechanism ensures that the digital twin model can accurately track the actual flood evolution process, promptly correct the simulation trajectory, and improve forecast accuracy and reliability.

[0095] After the simulation is completed, the model outputs a dynamic parameter set for the flood peak evolution, including key indicators such as water level and flow process curves at various control sections in different future time periods, flood propagation time, maximum inundation area, and water depth distribution. The parameters not only include expected values ​​but also probability distribution information, quantifying the uncertainty of the simulation results. The output data is encapsulated in a standardized format, including metadata such as timestamps, spatial locations, parameter types, numerical values, and confidence levels, facilitating parsing and use by downstream modules.

[0096] The digital twin model is tightly coupled with the visualization module, and the simulation results directly drive the dynamic presentation of the 3D scene. The model output interface converts abstract numerical results into graphical instructions that the graphics engine can parse, including visualization elements such as mesh vertex coordinates, color mapping relationships, and animation keyframes. Simulation parameters are transmitted to the rendering engine in real time via an application programming interface (API), dynamically generating flood inundation animations on the digital elevation model base map, intuitively demonstrating the entire flood evolution process. The visualization process supports multi-level rendering, displaying the global flood situation in the distance and showcasing local flow field details in the foreground, meeting the observation needs at different scales.

[0097] User interaction is a crucial feature of digital twin models. Through a visual interface, users can adjust initial conditions, modify engineering parameters, and set different scheduling schemes. The model responds to input in real time and recalculates. Interaction commands are captured through an event listening mechanism, parsed, and transformed into parameter adjustment commands that the model can recognize. The model employs an incremental calculation strategy, performing partial updates based on the original simulation results to quickly generate simulation results under the new scheme, ensuring the real-time nature and smoothness of the interaction.

[0098] The digital twin model also features a simulation playback function. All historical simulation states are recorded in a time-series database, forming a complete historical simulation trajectory. When the user clicks the playback control on the interface, the model extracts the simulation state sequence for the specified time period from the database and re-executes the simulation process at a set speed. During playback, auxiliary information such as water level curves and flow rate curves are displayed simultaneously to help users analyze key nodes and turning points in the simulation process and gain a deeper understanding of the flood evolution pattern.

[0099] Finally, the digital twin model transmits the simulation results to the multi-objective optimization module, providing scenario input for scheduling decisions. The output data includes complete scenario description information, encompassing not only hydrological and hydraulic parameters but also metadata such as constraint fulfillment during the simulation process and records of boundary condition changes. This rich information provides ample basis for subsequent optimization calculations, ensuring that the generated scheduling scheme not only meets the optimization objectives but also satisfies physical laws and engineering constraints, forming a complete closed loop from scenario simulation to decision optimization.

Claims

1. An intelligent supervision method based on a data visualization platform, characterized in that, Comprise: Step 1: collect meteorological radar data, river water level sensing data, soil moisture monitoring data and satellite remote sensing image, double-channel redundant check output time alignment initial data set; Step 2: convert the initial data set into a standardized space-time matrix, including: containerized adaptive unified data format, pre-trained variational autoencoder to calculate reconstruction error, adversarial generation of scanning abnormal data points, output cleaned space-time matrix with confidence label; Step 3: receive the cleaned space-time matrix, store the matrix data according to the confidence threshold, data with confidence greater than or equal to the threshold is stored in the cache library, historical data is stored in the space-time index library, monitor real-time business scenario state and data storage location, reinforcement learning decision generates routing strategy table, extracts data fragments from the storage library according to the routing strategy table and distributes them to the message queue; Step 4: extract data fragments from the message queue, input parallel processing, parallel processing includes: space-time graph convolution calculation generates flood peak evolution prediction results, Bayesian network and long short term memory calculation outputs risk assessment value, when integrated voting confidence is lower than threshold, activate federal learning parameter update, output analysis results with reliability index; Step 5: receive the analysis results output by step 4, digital twin model loads pre-stored physical constraint parameters, deduces flood control scene according to analysis results, outputs flood peak evolution dynamic parameters, inputs flood peak evolution dynamic parameters into multi-objective optimization calculation, iteratively generates scheduling instruction set, generates pre-play animation data according to scheduling instruction set; Step 6: dynamically render loading scheduling instruction set and pre-play animation data, generate interactive visualization interface and associate causal reasoning path, receive user feedback instructions to trigger federal model parameter update. 2.The data visualization platform-based intelligent supervision method according to claim 1, characterized in that, The step 1 comprises: The main channel transmits the original data stream to the platform entrance, and the auxiliary channel transmits the feature hash value to the platform entrance at the same time; The platform entrance receives the main channel data and calculates the real-time hash value of the main channel data; Extract the feature hash value transmitted by the auxiliary channel, compare the real-time hash value with the feature hash value, if they are inconsistent, trigger the data source node to retransmit the main channel data until the hash values are consistent, and output the initial data set with integrity mark. 3.The smart supervision method based on data visualization platform of claim 1, wherein, The step 2 comprises: Receive the initial data set, containerize and adapt to convert the initial data set, parse the meteorological data to obtain the rainfall intensity space-time matrix, and perform band operation on the remote sensing data to obtain the elevation grid, integrate the rainfall intensity space-time matrix and the elevation grid into a format unified matrix; Input the format unified matrix into the variational autoencoder to calculate the reconstruction error, input the format unified matrix into the adversarial generation network to scan the abnormal data points and output the abnormal score, and merge the reconstruction error and the abnormal score to generate the data confidence label; Map the data confidence label to the format unified matrix, and output the cleaned space-time matrix with confidence label. 4.The smart supervision method based on data visualization platform of claim 1, wherein, The step 3 comprises: Receive the cleaned space-time matrix, store the data according to the confidence threshold, data with confidence greater than or equal to the threshold is stored in the cache library and the space-time index is established, the rest of the data is stored in the historical database; Based on the stored data location, read the location state, capture the real-time business scene coding and data magnitude, input the location state, business scene coding and data magnitude into the reinforcement learning decision; The reward function calculates the optimal routing action, generates the routing strategy table, extracts the data shards from the storage according to the routing strategy table, and distributes the data shards to the message queue. 5.The smart supervision method based on data visualization platform of claim 1, wherein, The step 4 includes: Receiving the data shards distributed in step 3, constructing a watershed topology graph based on the water conservancy facility relationship in the data shards, the nodes being water conservancy facility entities and the edges being water flow directions and distances; Inputting the watershed topology graph into the spatio-temporal graph convolution calculation to output the flood peak evolution prediction result; Inputting the flood peak evolution prediction result into the risk assessment to output the risk assessment value; When the integrated voting confidence of the risk assessment value is lower than the threshold, activating the federated learning parameter update; Integrating the flood peak evolution prediction result, the risk assessment value and the federated learning update signal to output the analysis result with reliability indicators. 6.The smart supervision method based on data visualization platform of claim 1, wherein, The step 5 includes: Receiving the analysis result output in step 4, the digital twin model loading the pre-stored physical constraint parameters, deducing the flood control scene according to the analysis result, and outputting the flood peak evolution dynamic parameters; Inputting the flood peak evolution dynamic parameters into the multi-objective optimization calculation to generate a gate opening sequence by chromosome coding; The fitness calculation synchronously evaluates the flood control safety and water supply security, iteratively optimizes the gate opening sequence, outputs the Pareto optimal scheduling scheme, and generates the pre-play animation data according to the Pareto optimal scheduling scheme. 7.The smart supervision method based on data visualization platform of claim 6, wherein, The step 5 includes: Receiving the initial gate opening sequence, and screening the parent chromosomes by tournament selection; Inputting the parent chromosomes into the simulated binary crossover to generate a child chromosome sequence; Dynamically adjusting the weight coefficients α, β and γ of the child chromosome sequence, updating the fitness calculation rule, iteratively optimizing the child chromosome sequence based on the updated rule, and outputting the Pareto optimal scheduling data; Generating personnel transfer path data based on the Pareto optimal scheduling data. 8.The smart supervision method based on data visualization platform of claim 1, wherein, The step 6 includes: Dynamically rendering and loading the scheduling instruction set and the pre-play animation data to generate an interactive visualization interface, and superimposing the flood peak risk heat map on the digital elevation model base map; The interactive visualization interface receives the user's clicking instruction on the high-risk area to trigger the playback of the digital twin model deduction process; The visualization interface displays the pre-plan path animation, the animation shows the iterative trajectory of the multi-objective optimization algorithm, the interactive visualization interface receives the user's correction instruction to generate a difference vector, encrypts and uploads the difference vector, and triggers the federated model parameter update. 9.The smart supervision method based on data visualization platform of claim 8, wherein, The step 6 also includes: Receiving the encrypted uploaded difference vector, federated aggregation analyzing the difference vector, and calculating the global model parameter update amount; Distributing the global model parameter update amount to independent processing; The independent processing updates the risk assessment model parameters to generate updated risk assessment values; Inputting the updated risk assessment values into dynamic rendering to refresh the risk heat map of the visualization interface in real time.

10. An intelligent supervision system based on a data visualization platform, applied to the intelligent supervision method based on the data visualization platform according to any one of claims 1-9, characterized in that, It includes: A data acquisition unit configured to acquire meteorological radar data, river water level sensing data, soil moisture monitoring data and satellite remote sensing images, and execute double-channel redundant verification to output an initial data set aligned in time; The data processing unit is configured to convert the initial dataset into a standardized spatiotemporal matrix, including: containerizing and adapting to a unified data format, calculating the reconstruction error using a pre-trained variational autoencoder, generating adversarial data to scan outlier data points, and outputting a cleaned spatiotemporal matrix with confidence labels. The data analysis unit is configured to receive the cleaned spatiotemporal matrix, store the matrix data in layers according to the confidence threshold, monitor the real-time business scenario status and data storage location, generate a routing strategy table through reinforcement learning decision-making, and extract data from the repository and distribute it to the message queue according to the routing strategy table. The parallel processing unit is configured to extract data shards from the message queue, perform spatiotemporal graph convolution calculations to generate flood peak evolution prediction results, calculate risk assessment values ​​using Bayesian networks and long short-term memory, activate federated learning parameter updates when the ensemble voting confidence is below a threshold, and output analysis results with reliability indicators. The generation unit is configured to receive analysis results, load pre-stored physical constraint parameters into the digital twin model to simulate flood control scenarios, output dynamic parameters of flood peak evolution, perform multi-objective optimization calculations to iteratively generate a scheduling instruction set, and generate pre-simulation animation data based on the scheduling instruction set. The optimization unit is configured to dynamically render and load the scheduling instruction set and pre-rendering animation data, generate an interactive visualization interface and associate it with the causal inference path, and receive user feedback instructions to trigger the update of federated model parameters.

Images