An event-driven intelligent reservoir operation management system and method

Abstract

This invention provides an event-driven intelligent reservoir operation management system and method, belonging to the field of intelligent reservoir operation management technology. The event-driven intelligent reservoir operation management system of this invention includes a real-time data warehouse module, an event center module, a knowledge graph module, a reservoir microservice module, and an intelligent voice robot module. The technical solution of this invention collects data from the real-time data warehouse module, establishes an event center module based on the collected data to provide event publishing and subscription functions, and constructs a reservoir knowledge graph centered on events by combining existing database data and text data. Under different event-driven conditions, the reservoir microservice module initiates corresponding tasks and notifications, and the intelligent voice robot module can proactively provide users with graphics and dialogues that match the usage scenario, enabling the reservoir to have autonomous characteristics and improving the intelligence level of reservoir operation management.

Description

Technical Field

[0001] This invention relates to the field of Internet of Things (IoT) technology, and more specifically, to an event-driven intelligent reservoir operation management system and method. Background Technology

[0002] A reservoir is a water conservancy project that integrates functions such as flood control and disaster reduction, water supply and irrigation, power generation and navigation, and ecological improvement. Its operation status will directly affect regional production and life.

[0003] Currently, reservoir-related operation and management platforms mainly serve as auxiliary tools. Users interact with the platform through a graphical interface by clicking or touching, resulting in a very limited interaction method. Furthermore, existing reservoir operation and management platforms passively rely on pre-designed rules and algorithms to display data information, lacking the ability to proactively provide information. Summary of the Invention

[0004] The problem addressed by this invention is how to improve the intelligence level of reservoir operation and management.

[0005] To address the aforementioned problems, this invention provides an event-driven intelligent reservoir operation and management system, comprising a real-time data warehouse module, an event center module, a knowledge graph module, a reservoir microservice module, and an intelligent voice robot module. The real-time data warehouse module collects data from a business database, a monitoring database, and external data sources, and generates a first driving event based on the data, which is then published to the event center module. The event center module parses the first driving event into the knowledge graph module and notifies the reservoir microservice module and the intelligent voice robot module of the event based on the first driving event. The knowledge graph module constructs a reservoir knowledge graph based on existing database data, text data, and the first driving event. The reservoir microservice module initiates corresponding tasks and notifications based on the first driving event. The intelligent voice robot module generates corresponding scenario-based voice dialogues and graphical interfaces based on the first driving event, and broadcasts the voice dialogues and displays the graphical interfaces through an intelligent front-end device.

[0006] Optionally, the real-time data warehouse module is further configured to synchronize the data to the data lake and perform calculations on the data using streaming computing to generate the first driving event.

[0007] Optionally, the event center module includes an event library, an event annotation platform, and a message queue. The event library is used to merge similar events under the same event scenario. The event annotation platform is used to annotate the type and attributes of events to provide training data for the machine learning model of the real-time data warehouse module. The message queue is used for event subscription and publishing.

[0008] Optionally, the event library includes event categories, event attributes, event relationship categories, and event fusion rules compiled using event storming. The event center module is also used to fuse multiple similar first driving events according to the event fusion rules.

[0009] Optionally, the knowledge graph module includes a static entity knowledge graph and a dynamic event graph. The static entity knowledge graph is obtained by processing existing database data and text data, and the dynamic event graph is obtained by parsing the first driving event.

[0010] Optionally, for the existing database data, the D2R transformation method is used to process the existing database data, and for the text data, the pre-trained language model of SpacyNLP is used to perform word segmentation, part-of-speech tagging, entity linking, named entity recognition, and entity relation extraction on the text data.

[0011] Optionally, the reservoir microservice module is used to provide organizational office services, monitoring and early warning services, forecasting and scheduling services, and inspection services. The reservoir microservice module is also used to publish the second driving events generated in the process of providing services to the event center module.

[0012] Optionally, the intelligent voice robot module includes a natural language understanding unit, a dialogue management unit, a natural language generation unit, and a dialogue annotation platform; the natural language understanding unit processes the speech recognition results, extracts the user's intent and intent slot filling, and classifies the intent using a multi-task transformer framework; the dialogue management unit tracks the dialogue state and learns the dialogue strategy; the natural language generation unit transcribes the actions output by the dialogue management unit; and the dialogue annotation platform records the human-computer dialogue process and annotates and modifies the accuracy of the dialogue intent and intent slot filling.

[0013] Optionally, the dialogue annotation platform is also used to generate standard and accurate dialogue text to train a language model.

[0014] The event-driven intelligent reservoir operation and management system described in this invention collects data from a real-time data warehouse module, establishes an event center module based on the collected data to provide event publishing and subscription functions, and constructs a reservoir knowledge graph centered on events by combining existing database data and text data. Under different event-driven conditions, the reservoir microservice module initiates corresponding tasks and notifications, and the intelligent voice robot module can proactively provide users with graphics and dialogues that match the usage scenario, enabling the reservoir to have autonomous characteristics, improving the intelligence level of reservoir operation and management, and greatly facilitating reservoir managers to find information, quickly locate problems, make overall decisions, and schedule operations through the interaction of voice and graphical interface.

[0015] This invention also provides an event-driven intelligent reservoir operation management method, applied to the aforementioned event-driven intelligent reservoir operation management system, comprising: collecting data through a real-time data warehouse module and publishing a first driving event corresponding to the data to an event center module; the event center module receiving the first driving event and notifying a reservoir microservice module and an intelligent voice robot module of the event; the reservoir microservice module initiating corresponding tasks and notifications based on the first driving event; the intelligent voice robot module generating a voice dialogue and a graphical interface corresponding to the scenario based on the first driving event, and broadcasting the voice dialogue and displaying the graphical interface through an intelligent front-end device; and the event center module parsing the first driving event into the knowledge graph module. Attached Figure Description

[0016] Figure 1 This is a schematic diagram of an event-driven intelligent reservoir operation and management system according to an embodiment of the present invention. Detailed Implementation

[0017] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0018] like Figure 1 As shown, this embodiment of the invention provides an event-driven intelligent reservoir operation and management system, including a real-time data warehouse module, an event center module, a knowledge graph module, a reservoir microservice module, and an intelligent voice robot module.

[0019] The real-time data warehouse module is used to collect data from business databases, monitoring databases and external data sources, and generate a first driving event based on the data and publish it to the event center module.

[0020] Specifically, the data in the real-time data warehouse module mainly comes from business databases, monitoring databases, and external data sources. The business databases and monitoring databases contain structured data. Based on the metadata of the database tables, rules for triggering events are set. For example, when the water level field of the real-time water level table of a reservoir reaches the flood limit value, a reservoir water level warning event is generated. External data sources come from different sources, including both structured and unstructured data. Unstructured data includes disaster notification texts released by official social media. Structured data in external data sources refers to some formatted text data such as XML and JSON that are directly stored in the database. This data can be directly read and parsed by computers, such as the meteorological disaster data and the data returned by the interface for future refined rainfall prediction released by Caiyun Weather.

[0021] In reservoir application scenarios, the events generated by the real-time data warehouse module include water level warning events and rainfall warning events. Taking the water level warning event as the first driving event, a reservoir water level warning event (i.e., the first driving event) can be generated based on the data corresponding to the water level field of the real-time water level table reaching the flood limit water level value. Taking the rainfall warning event as an example, real-time rainfall data from the monitoring database rain gauge station can be collected, synchronized to the data lake, and then the streaming computing engine can calculate the data and issue a 24-hour rainfall warning event (i.e., the first driving event) to the event center module based on the fixed rule that the rainfall data exceeds 100 mm in 24 hours.

[0022] The real-time data warehouse module utilizes the real-time, low-latency characteristics of streaming computing to perform calculations and preprocessing on data (such as word segmentation and stop word removal for text data). It also combines rules and machine learning models to publish attributed events to the message queue of the event center (such as extracting specific events and attributes such as event classification, time of occurrence, and location from text data). The data before and after processing are persisted to the data lake in a hierarchical manner. The use of the streaming computing capabilities of the real-time data warehouse greatly reduces the latency of event processing.

[0023] The event center module is used to parse the first driving event into the knowledge graph module, and to notify the reservoir microservice module and the intelligent voice robot module of the event based on the first driving event.

[0024] Specifically, the event center module includes an event library, an event annotation platform, and a message queue. It extracts events in the reservoir domain to establish an event library and establishes an event center based on the message queue, providing event publishing and subscription functions.

[0025] The event database is used to fuse similar events within the same event scenario. Reservoir experts use event storming to organize event classifications and attributes, event relationship classifications, and event fusion rules to build the event database. The event relationship classifications include causal relationships, conditional relationships, inversion relationships, and sequential relationships. The event fusion rules use similarity algorithms such as Euclidean distance to fuse similar events (e.g., multiple similar primary driving events) within the same event scenario. Taking rainfall warning events as an example, multiple 24-hour rainfall warnings from multiple rain gauges in the reservoir basin that exceed 100 mm within a certain period can be fused into a single rainfall warning event, and the corresponding event can be parsed into a dynamic event graph.

[0026] Event storming is a flexible workshop format based on Domain-Driven Design (DDD) used for collaborative exploration of complex business domains. Domain experts and project team members brainstorm to list all domain events, then label each event with the command that caused it, and further label the role of the initiator of that command. Finally, the events are categorized into entities, aggregates, aggregate roots, and bounded contexts, and the domain model is built within the bounded context boundaries.

[0027] The event labeling platform is used to label the type and attributes of events, such as text and image data, to identify the event type and related attributes, thereby providing training data for the machine learning model of the real-time data warehouse module. For image labeling, manual identification is first performed; for example, if someone is swimming in a reservoir, a red box is drawn around the swimmer, and the event is labeled as illegal water activity.

[0028] For text annotations, such as the text regarding a reservoir water release event: "Following a notification from the Flood Control, Typhoon Prevention and Drought Relief Headquarters of City A, Reservoir B is scheduled to begin pre-release at 10:00 AM on July 22nd, opening its spillway at a flow rate of x cubic meters per second (excluding power generation and ecological purposes). The water level will be adjusted based on the subsequent typhoon impact and weather conditions after it drops below x meters. Downstream towns (townships) C, D, E, and F are requested to implement the water release warning work as required." The relevant attributes of the reservoir water release event include the reservoir name, start time, pre-release flow rate, and affected area. This is indicated by the annotations: "Reservoir Name" corresponds to "Reservoir B," "Start Time" corresponds to "10:00 AM on July 22nd," "Pre-release Flow Rate" corresponds to "x cubic meters per second," and "Affected Area" corresponds to "C, D, E, and F."

[0029] The message queue is used for event subscription and publishing. Events generated by reservoir operation and management are published to the event center. Subscribed event-driven systems and devices proactively perform tasks or generate voice dialogues and graphical interfaces based on the event scenario, giving the reservoir autonomous characteristics. In addition to completing some automated tasks, it can continuously learn and autonomously provide information to people, such as intelligent voice interaction, thereby meeting diverse needs. Under different event-driven conditions, it proactively displays graphics and dialogues that meet the needs of the actual scenario, greatly facilitating reservoir managers to find information, quickly locate problems, make overall decisions, and schedule operations.

[0030] The knowledge graph module is used to construct a reservoir knowledge graph based on existing database data, text data, and the first driving event.

[0031] Specifically, the knowledge graph module includes a static entity knowledge graph and a dynamic event graph. The static entity knowledge graph is obtained by processing existing database data and text data (derived from data achievements accumulated by users in previous information technology construction processes, which can be directly imported into the graph database, such as the database of all water conservancy project results left after the water conservancy census). For database data, the D2R (Database to RDF) transformation method is used, directly mapping database table names to classes in RDF (Resource Description Framework), fields to class attributes, and relationships between classes can be derived from the relationship table between tables. For text data, a pre-trained language model of SpacyNLP is used to perform operations such as word segmentation, part-of-speech tagging, entity linking, named entity recognition, and entity relationship extraction. For example, the engineering characteristics related to the reservoir are parsed from the document of the reservoir project scheduling and operation plan. The dynamic event graph is obtained by parsing the first driving event. The RDF data of the graph is finally persisted to the graph database. The model developed using natural language processing technology constructs a reservoir knowledge graph with events as its core, integrating basic static knowledge and dynamic events to support intelligent interaction.

[0032] The reservoir microservice module is used to initiate corresponding tasks and notifications based on the first driving event.

[0033] Specifically, the reservoir microservice module is used to provide organizational office services, monitoring and early warning services, forecasting and scheduling services, and inspection services. The reservoir microservice module is also used to publish the second driving events generated in the process of providing services to the event center module. That is, the events generated by the business are directly published to the event center module for processing, and the event center module is subscribed to start specific tasks and notifications.

[0034] The organizational services include notifications, pending approvals, contact lists, leave requests, and organizational structure; monitoring and early warning services include information on water levels exceeding warning levels, abnormal values ​​from dam safety monitoring equipment, and rainfall warnings; forecasting and dispatching services include analysis of reservoir rainfall carrying capacity, forecasting future flood conditions, and calculating feasible dispatching plans to cope with future floods; patrol services include assigning patrol tasks, recording patrol issues, and viewing patrol routes; specific tasks and notifications include: activation - real-time camera capture of water-related activities; activation - immediate notification to the administrator upon detection of water-related activities; activation - immediate notification to the administrator upon rainfall exceeding 50 mm within 30 minutes. For example, upon receiving a rainfall warning notification, the forecasting and dispatching service will initiate real-time flood tracking and forecasting.

[0035] The intelligent voice robot module is used to generate voice dialogue and graphical interface corresponding to the scene based on the first driving event, and to broadcast the voice dialogue and display the graphical interface through an intelligent front-end device.

[0036] Specifically, the intelligent voice robot module consists of Natural Language Understanding (NLU), Dialogue Management (DM), Natural Language Generation (NLG), and an annotation platform. It provides an interactive method combining voice and a graphical interface. The data generated from this interaction is annotated and refined, then fed back to the model for learning and training. Furthermore, it can be deeply integrated with existing business systems to achieve multi-terminal interaction across large screens, mobile devices, PCs, robot screens, and robot audio playback devices. For example, upon receiving a rainfall warning, the intelligent voice robot module broadcasts the warning through an intelligent front-end device and displays a graphical interface showing real-time flood tracking and forecasts, assisting users in scheduling and decision-making.

[0037] The NLU (Non-Logical Language Understanding) process processes the speech recognition results, extracting the user's intent and intent slot filling. A pre-trained language model using SpacyNLP is employed for text segmentation, text feature generation, and named entity recognition. A multi-task transformer framework (Dual Intent and Entity Transformer, DIET) is used to classify intents, addressing two key aspects of dialogue understanding: intent classification and entity recognition. DM (Dialogue Management) is divided into dialogue state tracking and dialogue policy learning, primarily updating the system state and generating corresponding actions based on the NLU's records. NLG (Non-Logical Language Learning) literalizes the actions output by DM, expressing the system's actions in text form. An annotation platform records the human-computer dialogue process, allowing for annotation and modification of the accuracy of the dialogue intent and intent slot filling. Standardized and accurate dialogue text is used to train the language model. For example: A user sends a voice message to the server, such as, "What is the water situation in the reservoir?"; the server converts the voice message to text, calls a model to analyze the text, identifies the intent name, such as, "Query reservoir water situation"; the required parameters for the intent, such as, time; returns a prompt when parameters are missing, such as, "When was the query time?"; the logic for processing the intent is then executed, such as, calling the water situation query API in the monitoring and early warning microservice; and the voice message is returned, such as: "The current water level in the reservoir is xx meters, and the reservoir capacity is xx million cubic meters."

[0038] Another embodiment of the present invention provides an event-driven intelligent reservoir operation management method, applied to the aforementioned event-driven intelligent reservoir operation management system, comprising: collecting data through a real-time data warehouse module and publishing a first driving event corresponding to the data to an event center module; the event center module receiving the first driving event and notifying a reservoir microservice module and an intelligent voice robot module of the event; the reservoir microservice module initiating corresponding tasks and notifications based on the first driving event; the intelligent voice robot module generating a voice dialogue and a graphical interface corresponding to the scenario based on the first driving event, and broadcasting the voice dialogue and displaying the graphical interface through an intelligent front-end device; and the event center module parsing the first driving event into the knowledge graph module.

[0039] While the present invention has been disclosed above, its scope of protection is not limited thereto. Those skilled in the art can make various changes and modifications without departing from the spirit and scope of the present invention, and all such changes and modifications will fall within the scope of protection of the present invention.

Claims

1. An event-driven intelligent reservoir operation and management system, characterized in that, It includes a real-time data warehouse module, an event center module, a knowledge graph module, a reservoir microservice module, and an intelligent voice robot module; The real-time data warehouse module is used to collect data from business databases, monitoring databases and external data sources, and generate a first driving event based on the data and publish it to the event center module. The event center module is used to parse the first driving event into the knowledge graph module, and to notify the reservoir microservice module and the intelligent voice robot module of the event based on the first driving event. The knowledge graph module is used to construct a reservoir knowledge graph based on existing database data, text data, and the first driving event. The reservoir microservice module is used to initiate corresponding tasks and notifications based on the first driving event; The intelligent voice robot module is used to generate voice dialogue and graphical interface corresponding to the scene based on the first driving event, and to broadcast the voice dialogue and display the graphical interface through an intelligent front-end device. The event center module includes an event library, an event annotation platform, and a message queue. The event library is used to fuse similar events in the same event scenario. The event annotation platform is used to annotate the type and attributes of events to provide training data for the machine learning model of the real-time data warehouse module. The message queue is used for event subscription and publishing. The event library includes event categories, event attributes, event relationship categories, and event fusion rules compiled using event storming. The event center module is also used to fuse multiple similar first driving events according to the event fusion rules.

2. The event-driven intelligent reservoir operation management system according to claim 1, characterized in that, The real-time data warehouse module is also used to synchronize the data to the data lake and to perform calculations on the data using streaming computing to generate the first driving event.

3. The event-driven intelligent reservoir operation management system according to claim 1, characterized in that, The knowledge graph module includes a static entity knowledge graph and a dynamic event graph. The static entity knowledge graph is obtained by processing existing database data and text data, and the dynamic event graph is obtained by parsing the first driving event.

4. The event-driven intelligent reservoir operation management system according to claim 3, characterized in that, For the existing database data, the D2R transformation method is used to process the existing database data. For the text data, the pre-trained language model of SpacyNLP is used to perform word segmentation, part-of-speech tagging, entity linking, named entity recognition, and entity relation extraction on the text data.

5. The event-driven intelligent reservoir operation management system according to claim 1, characterized in that, The reservoir microservice module is used to provide organizational office services, monitoring and early warning services, forecasting and scheduling services, and inspection services. The reservoir microservice module is also used to publish the second driving events generated in the process of providing services to the event center module.

6. The event-driven intelligent reservoir operation management system according to claim 1, characterized in that, The intelligent voice robot module includes a natural language understanding unit, a dialogue management unit, a natural language generation unit, and a dialogue annotation platform; The natural language understanding unit is used to process the results of speech recognition, extract the user's intent and intent slot filling, and classify the intent using a multi-task transformer framework; The dialogue management unit is used for dialogue state tracking and dialogue strategy learning; The natural language generation unit is used to textify the actions output by the dialogue management unit. The dialogue annotation platform is used to record the human-computer dialogue process and to annotate and modify the dialogue intent and whether the intent slot filling is accurate.

7. The event-driven intelligent reservoir operation management system according to claim 6, characterized in that, The dialogue annotation platform is also used to generate standard and accurate dialogue text for training language models.

8. An event-driven intelligent reservoir operation management method, applied to the event-driven intelligent reservoir operation management system according to any one of claims 1 to 7, characterized in that, include: Data is collected through the real-time data warehouse module, and the first driving event corresponding to the data is published to the event center module; The event center module receives the first driving event and notifies the reservoir microservice module and the intelligent voice robot module of the event. The reservoir microservice module initiates corresponding tasks and notifications based on the first driving event; The intelligent voice robot module generates a corresponding voice dialogue and graphical interface based on the first driving event, and broadcasts the voice dialogue and displays the graphical interface through an intelligent front-end device. The event center module parses the first driving event into the knowledge graph module.

Images