Dynamic visualization and intelligent alarm method and system based on multi-source water and rainfall condition monitoring data

Abstract

The application relates to the technical field of water intelligence, and particularly discloses a dynamic visualization and intelligent alarm method and system based on multi-source water and rainfall condition monitoring data. The method obtains multi-source real-time monitoring data such as water level, rainfall and water depth, combines an abnormal scoring model fusing statistical deviation and change rate to perform abnormal identification; in response to an abnormal event, a hierarchical alarm is triggered according to a severity level, and a dynamic visualization interface including a time sequence curve graph, a space heat map and an abnormal station list is automatically generated; meanwhile, based on the visualization content and the event type, a structured log draft is generated by a local lightweight language model. The system runs in a closed network environment, supports interactive chart linkage and local processing. The application realizes high-sensitivity detection, intuitive presentation and rapid response of hydrological abnormalities, significantly improves the efficiency of duty and the accuracy of decision-making, and guarantees data security.

Description

Technical Field

[0001] This invention relates to the field of intelligent water management technology, and in particular to a dynamic visualization and intelligent alarm method and system based on multi-source water and rainfall monitoring data. Background Technology

[0002] In the field of intelligent water management, traditional duty systems mainly rely on manual review of reports, static charts, or simple threshold alarms, which are insufficient to cope with complex and ever-changing water and rainfall conditions. While some existing systems have introduced automatic data acquisition and over-limit alarm functions, they generally suffer from problems such as a single anomaly detection mechanism, delayed response, and weak visualization capabilities. For example, most systems only use fixed thresholds to determine whether water levels exceed warning levels, failing to identify rapid water level jumps caused by short-duration heavy rainfall or trend anomalies caused by slow seepage. Alarm methods are mostly SMS or pop-ups, lacking tiered response and multi-channel collaboration, easily leading to missed alarms or excessive interference. Furthermore, current visualization modules typically display historical data in static chart form, lacking dynamic presentation capabilities such as spatiotemporal linkage, interactive drill-down, and predictive overlay, making it difficult for duty personnel to quickly grasp the overall risk. Although some research has attempted to introduce machine learning for anomaly detection, these models are often deployed in the cloud, posing a risk of data leakage, and are not integrated into local alarm and log reporting processes. Therefore, there is an urgent need for a hydrological monitoring method and system that integrates high-precision anomaly identification, hierarchical intelligent alarm, dynamic interactive visualization, and localized log assistance to improve the efficiency and intelligence of emergency response during the flood season, while meeting the requirements of water conservancy dedicated networks for data security and system autonomy and controllability. Summary of the Invention

[0003] The purpose of this invention is to provide a dynamic visualization and intelligent alarm method and system based on multi-source hydrological and rainfall monitoring data. This addresses the technical problems in existing technologies, such as reliance on fixed thresholds for hydrological data anomaly identification, difficulty in capturing non-stationary changes and trend risks, lack of tiered response and multi-channel collaboration in alarm mechanisms, static and isolated visualization presentation without support for interactive linkage and predictive overlay, disconnect between AI-generated content and business logs, untraceable processing, and excessive reliance on external networks leading to insufficient data security and localized control capabilities. By integrating a dynamic anomaly scoring model, tiered alarm strategies, multi-graph linkage visualization, and a local lightweight language generation mechanism, this invention achieves highly sensitive awareness, intuitive presentation, closed-loop response, and secure controllability of hydrological risks, effectively improving the level of intelligent water management.

[0004] To achieve the above objectives, a dynamic visualization and intelligent alarm method based on multi-source water and rainfall monitoring data is provided in a first aspect of the present invention, comprising: Acquire real-time monitoring data from multiple hydrological monitoring devices, wherein the real-time monitoring data includes at least water level data, rainfall data, and water depth data; The real-time monitoring data is subjected to anomaly pattern recognition to determine whether there is a data sequence that meets preset anomaly conditions. The preset anomaly conditions include water level exceeding a threshold, rainfall intensity increasing by more than a set value per unit time, monitoring values ​​undergoing sudden changes, or the duration of the abnormal state exceeding a set duration. In response to identifying a data sequence that meets the preset abnormal conditions, an alarm event is generated, and a corresponding alarm action is triggered based on the severity level of the alarm event. The alarm action includes sending a notification message to a designated terminal, initiating a voice call, activating an audio-visual prompt device, and highlighting the relevant area in the graphical user interface. Based on the real-time monitoring data and the alarm events, a dynamic visualization interface is automatically generated, which includes a time series curve, a spatial distribution heat map, and a list of abnormal monitoring stations. Chart metadata describing the structure and attributes of the above charts is also generated. The time series curve displays warning line reference values ​​overlaid, and the spatial distribution heat map colors the monitoring values ​​according to geographical regions. Based on the content presented by the dynamic visualization interface, the chart metadata, and the type of the alarm event, the language generation model outputs a structured text description, and fills the structured text description into the log recording template to form an editable log draft.

[0005] Furthermore, anomaly pattern identification of real-time monitoring data includes: calculating anomaly scores for the current monitoring data. When the abnormal score When the threshold is exceeded, a data sequence is determined to meet preset abnormal conditions; wherein, the abnormality score Determine using the following formula: .

[0006] Indicates the current time The monitoring values; and These represent the reference mean and reference standard deviation, respectively, calculated based on historical data windows from the same period. This represents the rate of change of a value within the most recent time window; This indicates the duration the system has been in a stable state since the last abnormal event occurred; The preset attenuation coefficient has a value range of 0.1 to 0.5. and For dynamic weighting coefficients, satisfying ,and It increases as the current rainfall intensity level increases.

[0007] Furthermore, the dynamic visualization interface supports user interaction. In response to the user's selection of a region in the heat map, it automatically updates the time series curve to display the historical data of all monitoring points in that region. In response to the user's selection of an item in the list of abnormal stations, it synchronously marks the location of the abnormal station in the heat map and enlarges its trend curve.

[0008] Furthermore, the severity level of the alarm event is determined comprehensively based on the degree to which the abnormal value deviates from the threshold, the duration, and the scope of impact, and is mapped to a three-level alarm strategy: The first level only pushes messages to the terminal; The second level triggers a voice call while pushing a message; The third level additionally activates the locally deployed audio-visual prompts, building upon the second level.

[0009] Furthermore, the language generation model is a lightweight neural network model deployed on a local computing device. Its inputs include the timestamp of the abnormal event, the identifier of the monitoring point involved, the abnormal type label, and the metadata of the corresponding visualization chart. The output is a natural language paragraph conforming to a preset format.

[0010] Secondly, the present invention also provides a dynamic visualization and intelligent alarm system based on multi-source water and rainfall monitoring data, applied to any of the dynamic visualization and intelligent alarm methods based on multi-source water and rainfall monitoring data described in any one of the claims, comprising: The data acquisition unit is configured to acquire real-time monitoring data from multiple hydrological monitoring devices; An anomaly detection unit is configured to perform anomaly pattern recognition on the real-time monitoring data to determine whether there is a data sequence that meets preset anomaly conditions; The alarm execution unit, in response to the judgment result of the anomaly detection unit, executes the corresponding alarm action according to the severity level of the alarm event; The visualization generation unit is configured to generate a dynamic visualization interface based on the real-time monitoring data and alarm events, including time series curves, spatial distribution heat maps, and a list of abnormal monitoring stations. The log auxiliary unit is configured to call a language generation model to generate a structured text description based on the content of the dynamic visualization interface, and then fill it into the log recording template to form an editable log draft.

[0011] Furthermore, the visualization generation unit is integrated into the client interface of the hydrological business management system, presented as a movable and foldable floating window, and automatically switches to a semi-transparent state when there is no user interaction within a preset time.

[0012] Furthermore, the data acquisition unit integrates an adaptive transmission control subunit for managing the data flow strategy between the data acquisition unit and the anomaly detection unit. The adaptive transmission control subunit is configured to monitor the network communication status in real time: when the network communication status is normal, it controls the data acquisition unit to transmit the real-time monitoring data in full to the anomaly detection unit for centralized processing; When a network communication interruption or delay exceeding a preset threshold is detected, the system automatically switches to edge caching mode, instructs the data acquisition unit to establish a data cache queue with the most recent time window in the local storage area, performs preliminary screening on the data cache queue according to preset detection logic, marks compressed data packets and corresponding alarm events that are determined to be abnormal as high priority, and synchronizes them to the anomaly detection unit through a breakpoint resume mechanism.

[0013] Furthermore, the alarm execution unit includes a communication interface subunit and a peripheral control subunit. The communication interface subunit is connected to the SMS gateway and the voice call service, and the peripheral control subunit is electrically connected to the audio-visual prompt device for receiving control commands and driving it to work.

[0014] Thirdly, the present invention also provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the dynamic visualization and intelligent alarm method based on multi-source water and rainfall monitoring data as described in any one of the claims.

[0015] The beneficial technical effects of the present invention are at least as follows: This invention provides a dynamic visualization and intelligent alarm method and system based on multi-source water and rainfall monitoring data, which has the following outstanding advantages: Firstly, by fusing and analyzing real-time monitoring data from multiple sources such as water level, rainfall, and water depth, and introducing a dynamic anomaly scoring model that incorporates statistical bias, instantaneous change rate, and stability decay factor, the sensitivity and accuracy of identifying complex anomaly patterns such as water levels exceeding warning levels, short-duration heavy rainfall, and data jumps are significantly improved. This mechanism breaks through the limitations of traditional fixed threshold alarms and effectively solves the problem of delayed response of existing systems to trend risks and sudden events.

[0016] Secondly, based on the severity level of alarm events, a graded response strategy (message push, voice call, and audio-visual prompts) is automatically triggered, and a dynamic visualization interface including time series curves, spatial heat maps, and a list of abnormal monitoring stations is generated simultaneously, supporting interactive drill-down and chart linkage. At the same time, a locally deployed lightweight language generation model automatically generates structured log drafts, with the content including data sources and event context, realizing an integrated closed loop of perception, alarm, presentation, and recording, which greatly improves the efficiency of on-duty personnel in judgment and the consistency of operation.

[0017] Third, the entire system operates in a closed network environment. Anomaly detection, visualization rendering, and text generation are all completed on local servers or edge devices, without relying on external public cloud services. All visualization snapshots and log drafts are associated with time indexes and encrypted for archiving, ensuring that the processing is traceable and the data does not leave the domain. Attached Figure Description

[0018] The present invention will be further described with reference to the accompanying drawings, but the embodiments in the drawings do not constitute any limitation on the present invention. For those skilled in the art, other drawings can be obtained based on the following drawings without creative effort.

[0019] Figure 1 This is a flowchart illustrating the working steps of a dynamic visualization and intelligent alarm method based on multi-source water and rainfall monitoring data, as disclosed in one embodiment of the present invention. Figure 2 This is a schematic diagram of a dynamic visualization and intelligent alarm system based on multi-source water and rainfall monitoring data disclosed in one embodiment of the present invention. Detailed Implementation

[0020] Embodiments of the present invention are described in detail below. Examples of these embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present invention, and should not be construed as limiting the present invention.

[0021] Example 1 refer to Figure 1 This invention provides an embodiment of a dynamic visualization and intelligent alarm method based on multi-source water and rainfall monitoring data, comprising: S1. Acquire real-time monitoring data from multiple hydrological monitoring devices, wherein the real-time monitoring data includes at least water level data, rainfall data, and water depth data; S2. Perform anomaly pattern recognition on the real-time monitoring data to determine whether there is a data sequence that meets preset anomaly conditions. The preset anomaly conditions include water level exceeding a threshold, rainfall intensity increasing by more than a set value per unit time, or a sudden change in the monitoring value. S3. In response to identifying a data sequence that meets the preset abnormal conditions, generate an alarm event and trigger a corresponding alarm action based on the severity level of the alarm event. The alarm action includes sending a notification message to a designated terminal, activating an audio-visual prompt device, and highlighting the relevant area in the graphical user interface. S4. Based on the real-time monitoring data and the alarm events, automatically generate a dynamic visualization interface that includes a time series curve, a spatial distribution heat map, and a list of abnormal monitoring stations. The time series curve displays warning line reference values ​​overlaid, and the spatial distribution heat map colors-codes the monitoring values ​​according to geographical regions. S5. Based on the content presented by the dynamic visualization interface and the type of the alarm event, the language generation model outputs a structured text description, and the structured text description is filled into the log recording template to form an editable log draft.

[0022] In this embodiment, the system is deployed on a local server at the city's water management center and connected to a citywide hydrological monitoring network, including river level stations, rain gauges, and road flooding sensors. The system acquires real-time data streams from each monitoring device at preset intervals (e.g., every 5 minutes). The data includes at least water level values ​​(in meters), hourly cumulative rainfall (in millimeters), and water depth (in centimeters). All data carries a unified timestamp and a unique identifier for the measuring point.

[0023] Upon receiving the data, the system first performs anomaly pattern identification. For water level data, it determines whether it exceeds a preset warning threshold; for rainfall data, it calculates the rate of change per unit time. If the increase exceeds a set slope threshold, it is considered a short-duration heavy rainfall event. Simultaneously, a sliding window analysis is used to analyze the numerical sequences of each measuring point to detect any abrupt changes (such as a water level rise exceeding 0.3 meters within 30 seconds). When any condition is met, the system generates an alarm event and comprehensively assesses its severity level based on the magnitude of the anomaly, its duration, and the number of affected measuring points.

[0024] Based on the alarm level, the system automatically triggers a tiered response: low-level events only push notifications to the mobile terminals of on-duty personnel; medium-level events initiate voice calls in addition to the push notifications; and high-level events further activate the audio-visual prompts in the duty room and mark the affected area in red with a flashing highlight in the graphical user interface of the business management system.

[0025] Meanwhile, the visualization generation module automatically constructs a dynamic visualization interface, including: (1) a time series curve, with the horizontal axis representing time and the vertical axis representing the monitored value, overlaid with blue warning lines and red guarantee lines; (2) a spatial distribution heat map, based on the GIS base map, rendering the rainfall intensity or water level value by administrative region or watershed unit with color gradient; (3) a list of abnormal stations, displaying the names, current values, and deviations of abnormal stations in risk order. The above views support interactive linkage—clicking on a region in the heat map automatically switches the curve to the historical trend of all monitoring points in that region; clicking on a station in the list focuses the heat map and zooms in on its location.

[0026] Finally, the log auxiliary module calls the locally deployed lightweight language generation model, taking into account alarm type, involved measurement points, anomaly characteristics, and metadata of visualization charts, and outputting a structured text, such as: "The water level of the XX river section exceeded the warning level at 14:20 today, currently reaching 5.8 meters, 0.3 meters higher than the warning line. Three surrounding water accumulation points are also showing a rising water depth trend." This text is automatically filled into the standard duty log template, forming an editable draft, which can be reviewed and confirmed by the duty personnel before being submitted and archived with one click.

[0027] The entire process is completed within a closed network, without accessing external services, ensuring data security and real-time system response.

[0028] Furthermore, anomaly pattern identification of real-time monitoring data includes: calculating anomaly scores for the current monitoring data. When the abnormal score When the threshold is exceeded, a data sequence is determined to meet preset abnormal conditions; wherein, the abnormality score Determine using the following formula: .

[0029] Indicates the current time The monitoring values; and These represent the reference mean and reference standard deviation, respectively, calculated based on historical data windows from the same period. This represents the rate of change of a value within the most recent time window; This indicates the duration the system has been in a stable state since the last abnormal event occurred; The preset attenuation coefficient has a value range of 0.1 to 0.5. and For dynamic weighting coefficients, satisfying ,and It increases as the current rainfall intensity level increases.

[0030] Furthermore, the dynamic visualization interface supports user interaction. In response to the user's selection of a region in the heat map, it automatically updates the time series curve to display the historical data of all monitoring points in that region. In response to the user's selection of an item in the list of abnormal stations, it synchronously marks the location of the abnormal station in the heat map and enlarges its trend curve.

[0031] In this embodiment, the dynamic visualization interface is implemented using a web-based front-end framework (such as Vue.js combined with ECharts and Leaflet), supporting rich user interaction operations. When a user selects a geographical area (e.g., an administrative region, a watershed subunit, or a custom polygon range) on the spatial distribution heatmap by clicking or touching, the front-end interface immediately sends a region query request to the back-end visualization service module. Based on the spatial boundary of the region, the back-end retrieves historical data of all hydrological monitoring points falling within the range from the time-series database, and returns the results after aligning them by time. The front-end then automatically refreshes the time-series curve, overlays and plots the water level or rainfall change curves of all monitoring points in the selected area, distinguishes each monitoring point with different colors, and retains the warning line reference value as a background benchmark, making it convenient for users to horizontally compare the response consistency of each station within the region.

[0032] On the other hand, when a user clicks on an entry in the list of abnormal monitoring stations (such as "Station A: Sudden Increase in Water Level"), the system first locates the station's geographical coordinates on the heatmap and adds a highlighted marker (such as a flashing red circle or an icon pop-up) at its location. Simultaneously, the time-series graph automatically switches to display only the station's historical data and enlarges its trend window for the most recent 24 hours, highlighting the details of changes before and after the anomaly. If the station is associated with multiple monitoring indicators (such as water level and water depth), the graph can switch or overlay multi-dimensional data. All of the above interactions are implemented through lightweight communication between the front-end event listener and the back-end API, with response latency controlled to the millisecond level, ensuring smooth operation and information synchronization, and significantly improving the ability of on-duty personnel to quickly assess local risks.

[0033] Furthermore, the severity level of the alarm event is determined comprehensively based on the degree to which the abnormal value deviates from the threshold, the duration, and the scope of impact, and is mapped to a three-level alarm strategy: The first level only pushes messages to the terminal; The second level triggers a voice call while pushing a message; The third level additionally activates the locally deployed audio-visual prompts, building upon the second level.

[0034] In this embodiment, the severity level of the alarm event is dynamically calculated and determined by the alarm assessment module. After identifying the abnormal data sequence, the module generates a comprehensive risk score by combining the degree of deviation of the abnormal value from the threshold, the duration, and the scope of impact, and maps it to the three-level alarm strategy accordingly.

[0035] When an alarm is identified as Level 1, the system pushes a structured notification to the terminal of the on-duty personnel and simultaneously sends a lightweight inspection command to the AI ​​on-duty robot deployed in the water management center. The robot parses the command based on the natural language understanding module, automatically retrieves the historical records and equipment status of relevant monitoring stations in the business system, generates a preliminary analysis summary, and presents it in the form of a dialogue in the sidebar of the visualization interface to help the operator quickly grasp the background information.

[0036] If the alarm is upgraded to Level 2, in addition to push notifications and voice calls, the system instructs the AI ​​duty robot to proactively initiate interactive inquiries, such as: "The water level in section A of the river has exceeded the warning level. Please confirm whether to activate Level 3 emergency response?" and provide quick options such as "Yes / No / View Details". Duty personnel can respond directly by voice or click, and the robot will then execute subsequent procedures (such as retrieving emergency plans and notifying rescue teams), achieving human-machine collaborative decision-making.

[0037] When a Level 3 alarm is triggered, the system activates the audible and visual alerts and authorizes the AI-powered duty robot to perform high-priority automated operations. These operations include automatically generating emergency reports, pushing them to the command center screen, linking with the video surveillance system to retrieve real-time footage, and sending emergency emails with attachments to relevant personnel. The robot can also, according to preset rules, shut off water supply valves in the affected area or adjust pump station operating parameters without human intervention (operation permissions must be pre-configured). All robot actions are logged and linked to alarm events to ensure auditability and traceability.

[0038] By deeply integrating the tiered alarm mechanism with the AI ​​duty robot, not only has the intelligence level of alarm response been improved, but a business closed loop has also been realized from "passive notification" to "proactive handling," significantly enhancing the resilience and emergency response efficiency of the urban water system.

[0039] Furthermore, the language generation model is a lightweight neural network model deployed on a local computing device. Its inputs include the timestamp of the abnormal event, the identifier of the monitoring point involved, the abnormal type label, and the metadata of the corresponding visualization chart. The output is a natural language paragraph conforming to a preset format.

[0040] refer to Figure 2 Secondly, the present invention also provides a dynamic visualization and intelligent alarm system based on multi-source water and rainfall monitoring data, applied to any of the dynamic visualization and intelligent alarm methods based on multi-source water and rainfall monitoring data described in any one of the claims, comprising: The data acquisition unit is configured to acquire real-time monitoring data from multiple hydrological monitoring devices; An anomaly detection unit is configured to perform anomaly pattern recognition on the real-time monitoring data to determine whether there is a data sequence that meets preset anomaly conditions; The alarm execution unit, in response to the judgment result of the anomaly detection unit, executes the corresponding alarm action according to the severity level of the alarm event; The visualization generation unit is configured to generate a dynamic visualization interface based on the real-time monitoring data and alarm events, including time series curves, spatial distribution heat maps, and a list of abnormal monitoring stations. The log auxiliary unit is configured to call a language generation model to generate a structured text description based on the content of the dynamic visualization interface, and then fill it into the log recording template to form an editable log draft.

[0041] Furthermore, the visualization generation unit is integrated into the client interface of the hydrological business management system, presented as a movable and foldable floating window, and automatically switches to a semi-transparent state when there is no user interaction within a preset time.

[0042] Furthermore, the visualization generation unit is integrated into the client interface of the hydrological business management system, presented as a movable and foldable floating window, and automatically switches to a semi-transparent state when there is no user interaction within a preset time.

[0043] Furthermore, the alarm execution unit includes a communication interface subunit and a peripheral control subunit. The communication interface subunit is connected to the SMS gateway and the voice call service, and the peripheral control subunit is electrically connected to the audio-visual prompt device for receiving control commands and driving it to work.

[0044] Thirdly, the present invention also provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the dynamic visualization and intelligent alarm method based on multi-source water and rainfall monitoring data as described in any one of the claims.

[0045] This embodiment provides a dynamic visualization and intelligent alarm method, system, and storage medium based on multi-source hydrological and rainfall monitoring data. By aggregating heterogeneous data streams from various types of hydrological monitoring equipment, such as river level stations, rain gauges, and water accumulation point sensors, in real time, and performing unified timestamp alignment, unit standardization, field semantic mapping, and metadata annotation on the raw data, a standard hydrological dataset with standardized format, spatiotemporal consistency, and traceable source is constructed. This standardized governance mechanism effectively solves the information fragmentation problem caused by inconsistent equipment protocols, data granularity differences, and semantic ambiguity in traditional hydrological duty systems, laying a high-quality data foundation for subsequent anomaly identification, visualization, and intelligent log generation.

[0046] Leveraging a locally deployed hierarchical alarm mechanism and dynamic visualization engine, the system automatically determines event levels and triggers corresponding alarm actions based on an anomaly scoring model that integrates statistical deviation, rate of change, and stability decay factors. Simultaneously, it generates an interactive visualization interface including time-series curves with overlaid warning lines, geospatial heat maps, and a list of abnormal monitoring stations, supporting user drill-down and chart linkage. Furthermore, a lightweight language generation model, combined with visualization content and event context, automatically generates structured log drafts and links them to original chart snapshots, achieving a closed-loop process from data perception to alarm response and record archiving. This significantly reduces the burden of manual intervention and improves the standardization and efficiency of shift operations.

[0047] The entire system operates in a closed network environment, with all AI inference, visualization rendering, and data processing completed on local servers or edge devices, without relying on external public cloud services. All operational behaviors, alarm records, and log versions are encrypted and stored with a time index, supporting full lifecycle auditing and traceability. This solution, while ensuring data security and system autonomy and controllability, achieves highly sensitive identification, intuitive presentation, and rapid response to hydrological risks, demonstrating strong engineering feasibility and industry promotion value.

[0048] In the description of this specification, the references to terms such as "an embodiment," "example," "specific example," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the present invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0049] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit. If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes multiple instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing programs, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

Claims

1. A dynamic visualization and intelligent alarm method based on multi-source water and rainfall monitoring data, characterized in that, include: Acquire real-time monitoring data from multiple hydrological monitoring devices, wherein the real-time monitoring data includes at least water level data, rainfall data, and water depth data; The real-time monitoring data is subjected to anomaly pattern recognition to determine whether there is a data sequence that meets preset anomaly conditions, including: calculating the anomaly score of the current monitoring data. When the abnormal score When the threshold is exceeded, a data sequence is determined to meet preset abnormal conditions; wherein, the abnormality score Determine using the following formula: ； Indicates the current time The monitoring values; and These represent the reference mean and reference standard deviation, respectively, calculated based on historical data windows from the same period. This represents the rate of change of a value within the most recent time window; This indicates the duration the system has been in a stable state since the last abnormal event occurred; The preset attenuation coefficient has a value range of 0.1 to 0.

5. and For dynamic weighting coefficients, satisfying ,and The intensity increases with the current rainfall intensity level; in response to identifying a data sequence that meets the preset abnormal conditions, an alarm event is generated, and a corresponding alarm action is triggered based on the severity level of the alarm event; Based on the real-time monitoring data and the alarm events, a dynamic visualization interface is automatically generated, which includes a time series curve, a spatial distribution heat map, and a list of abnormal monitoring stations, and the chart metadata is output; wherein the time series curve displays the warning line reference value overlaid, and the spatial distribution heat map color-codes the monitoring values ​​according to geographical regions; Based on the dynamic visualization interface, the chart metadata, and the type of the alarm event, a structured text description is output according to the language generation model, and the structured text description is filled into the log recording template.

2. The dynamic visualization and intelligent alarm method based on multi-source water and rainfall monitoring data according to claim 1, characterized in that, The dynamic visualization interface supports user interaction. In response to the user's selection of a region in the heat map, it automatically updates the time series curve to display the historical data of all monitoring points in that region. In response to the user's selection of an item in the list of abnormal stations, it synchronously marks the location of the abnormal station in the heat map and enlarges its trend curve.

3. The dynamic visualization and intelligent alarm method based on multi-source water and rainfall monitoring data according to claim 1, characterized in that, The severity level of the alarm event is determined comprehensively based on the degree to which the abnormal value deviates from the threshold, the duration, and the scope of impact, and is mapped to a three-level alarm strategy: The first level only pushes messages to the terminal; The second level triggers a voice call while pushing a message; The third level additionally activates the locally deployed audio-visual prompts, building upon the second level.

4. The dynamic visualization and intelligent alarm method based on multi-source water and rainfall monitoring data according to claim 1, characterized in that, The language generation model is a lightweight neural network model deployed on a local computing device. Its inputs include the timestamp of the abnormal event, the identifier of the monitoring point involved, the abnormal type label, and the metadata of the corresponding visualization chart. The output is a natural language paragraph that conforms to a preset format.

5. A dynamic visualization and intelligent alarm system based on multi-source water and rainfall monitoring data, applied to the dynamic visualization and intelligent alarm method based on multi-source water and rainfall monitoring data as described in any one of claims 1-4, characterized in that, include: The data acquisition unit is configured to acquire real-time monitoring data from multiple hydrological monitoring devices; An anomaly detection unit is configured to perform anomaly pattern recognition on the real-time monitoring data to determine whether there is a data sequence that meets preset anomaly conditions, including: calculating anomaly scores for the current monitoring data. When the abnormal score When the threshold is exceeded, a data sequence is determined to meet preset abnormal conditions; wherein, the abnormality score Determine using the following formula: ； Indicates the current time The monitoring values; and These represent the reference mean and reference standard deviation, respectively, calculated based on historical data windows from the same period. This represents the rate of change of a value within the most recent time window; This indicates the duration the system has been in a stable state since the last abnormal event occurred; The preset attenuation coefficient has a value range of 0.1 to 0.

5. and For dynamic weighting coefficients, satisfying ,and It increases as the current rainfall intensity level increases; The alarm execution unit, in response to the judgment result of the anomaly detection unit, executes the corresponding alarm action according to the severity level of the alarm event; The visualization generation unit is configured to generate a dynamic visualization interface based on the real-time monitoring data and alarm events, including time series curves, spatial distribution heat maps, and a list of abnormal monitoring stations. The log auxiliary unit is configured to call a language generation model to generate a structured text description based on the content of the dynamic visualization interface, and then fill it into the log recording template to form an editable log draft.

6. The dynamic visualization and intelligent alarm system based on multi-source water and rainfall monitoring data according to claim 5, characterized in that, The visualization generation unit is integrated into the client interface of the hydrological business management system. It is presented as a movable and foldable floating window, and automatically switches to a semi-transparent state when there is no user interaction within a preset time.

7. The dynamic visualization and intelligent alarm system based on multi-source water and rainfall monitoring data according to claim 5, characterized in that, The data acquisition unit integrates an adaptive transmission control subunit, which is used to manage the data flow strategy between the data acquisition unit and the anomaly detection unit. The adaptive transmission control subunit is configured to monitor the network communication status in real time: when the network communication status is normal, it controls the data acquisition unit to transmit the real-time monitoring data in full to the anomaly detection unit for centralized processing; When a network communication interruption or delay exceeding a preset threshold is detected, the system automatically switches to edge caching mode, instructs the data acquisition unit to establish a data cache queue with the most recent time window in the local storage area, performs preliminary screening on the data cache queue according to preset detection logic, marks compressed data packets and corresponding alarm events that are determined to be abnormal as high priority, and synchronizes them to the anomaly detection unit through a breakpoint resume mechanism.

8. The dynamic visualization and intelligent alarm system based on multi-source water and rainfall monitoring data according to claim 5, characterized in that, The alarm execution unit includes a communication interface subunit and a peripheral control subunit. The communication interface subunit is connected to the SMS gateway and the voice call service. The peripheral control subunit is electrically connected to the sound and light prompt device and is used to receive control commands and drive it to work.

9. A computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the dynamic visualization and intelligent alarm method based on multi-source water and rainfall monitoring data as described in any one of claims 1 to 4.

Images