An event element automatic abstract generation method and system based on a large language model

Abstract

The application provides an event element automatic abstract generation method and system based on a large language model, and belongs to the technical field of environmental monitoring. The application obtains abnormal discharge, water quality and operation state records of a water supply and drainage pipe network in a target period; determines an effective time period, a connection relationship, an isolation relationship and a medium conveying direction according to the operation state records, and arranges event records into a water supply event record sequence and a drainage event record sequence; generates an event element generation instruction including a pollution source, a pollution invasion site, an affected water supply object, a disposal matter and a corresponding event record identifier for the water supply and drainage connection relationship; inputs the event record sequence and the event element generation instruction into a large language model to generate a cross-contamination event element, and inputs the cross-contamination event elements of the same water supply and drainage connection relationship into the large language model to generate a cross-contamination event abstract. The application can improve the association integrity of monitoring information and operation and maintenance events and the abstract accuracy.

Description

Technical Field

[0001] This application belongs to the field of environmental monitoring technology, and in particular relates to a method and system for automatic event element summarization based on a large language model. Background Technology Large language models possess the ability to understand event records, summarize event elements, and generate summary text. They can be used to organize scattered operational information into structured event elements and event summaries. In urban underground water supply, drainage, and pipeline network operation and maintenance scenarios, combining water quality monitoring, emission monitoring, and pipeline operation records to generate event summaries helps to reflect the event process between abnormal emissions, water quality alarms, and response activities.

[0002] Existing methods for automatic event element summarization based on large language models typically extract core events from the original event text and construct a model input by combining event trigger words, prompts, alarm information, or monitoring data. The large language model then identifies event elements and generates an event summary. Environmental monitoring data typically includes water quality monitoring results from water supply networks, discharge monitoring results from drainage outlets, and abnormal alarm records.

[0003] However, environmental monitoring alarms typically only reflect abnormalities at specific monitoring locations and times. Changes in valve opening / closing, isolation operations, and media transport status within water supply and drainage networks can alter the network objects and scope of impact affected by the abnormal media. When processing event text and environmental monitoring data in isolation, event summaries struggle to accurately correlate abnormal emission sources, abnormal water supply locations, affected objects, and isolation and treatment activities. Therefore, existing technologies suffer from incomplete correlation between environmental monitoring information and operational event elements in cross-contamination incidents related to water supply and drainage, and insufficient accuracy in event summary content. Summary of the Invention

[0004] The purpose of this application is to provide a method, system, and electronic device for automatically generating event element summaries based on a large language model, in order to solve the problems of incomplete element association and insufficient accuracy of event summary content in existing technologies for cross-contamination events involving water supply and drainage.

[0005] To address the aforementioned technical problems, in a first aspect, this application provides a method for automatically generating event element summaries based on a large language model, comprising: Acquire event records and operational status records of the water supply and drainage network within the target time period. Event records include event record identifier, record time, network object identifier, and record content. Based on the operational status records, determine the operational topology status, including the effective time period, connectivity, isolation, and media delivery direction; For each operating topology state, events whose recording time falls within the valid time period are arranged into a water supply event record sequence and a drainage event record sequence according to the pipeline object identifier, connectivity relationship and media delivery direction; For the water supply and drainage connectivity relationship between the water supply event record sequence and the drainage event record sequence, an event element generation instruction is generated according to the medium transport direction and isolation relationship. The event element generation instruction includes the pollution source, the pollutant intrusion site, the affected water supply object, the handling matters, and the corresponding event record identifier. Input the water supply event record sequence, drainage event record sequence, and event element generation instructions into the big language model to generate cross-contamination event elements that include element categories, event element content, and corresponding event record identifiers; Input the cross-contamination event elements corresponding to the same water supply and drainage connection relationship into the large language model to generate a cross-contamination event summary.

[0006] In one feasible implementation, based on the operational status record, the operational topology status, including effective time periods, connectivity relationships, isolation relationships, and media transport direction, is determined, including: Based on the status record time in the operation status record, the pipeline network object identification at both ends of the valve, and the valve opening and closing status, determine the status change time when the valve opening and closing status changes, and divide the time period into time periods based on the start and end time of the target time period and the status change time. For each time period, based on the valve opening and closing status corresponding to the start time of the time period in the operation status record, for the pipeline objects at both ends of the valve in the open state, the connection relationship and the direction of medium transportation are determined according to the pipeline object identifiers of the starting point and the ending point of medium transportation; for the pipeline objects at both ends of the valve in the closed state, the isolation relationship is determined. Each time period is defined as a valid time period, and the valid time period, the corresponding connectivity, isolation, and media transport direction are defined as an operating topology state.

[0007] In one feasible implementation, for each operating topology state, event records whose recording time falls within the valid time period are arranged into a water supply event record sequence and a drainage event record sequence according to the pipeline object identifier, connectivity, and media transport direction, including: Event records whose recording time is within the valid time period are divided into a water supply record set and a drainage record set. The water supply record set includes event records of water supply pipe network objects corresponding to pipe network object identifiers, and the drainage record set includes event records of drainage pipe network objects corresponding to pipe network object identifiers. According to the connectivity, the water supply network object identifiers in the water supply record set and the drainage network object identifiers in the drainage record set are arranged sequentially along the medium transport direction to obtain the water supply network object arrangement and the drainage network object arrangement. Arrange the event records in the water supply record set and the drainage record set according to the water supply network object arrangement and drainage network object arrangement respectively. Then, arrange the event records with the same network object identifier according to the recording time to obtain the water supply event record sequence and the drainage event record sequence.

[0008] In one feasible implementation, for the water supply and drainage connectivity relationship between the water supply event record sequence and the drainage event record sequence, an event element generation instruction is generated according to the medium transport direction and isolation relationship, including: From the water supply event record sequence and the drainage event record sequence, identify the water supply network object identifier, the drainage network object identifier, and the corresponding event record identifier that have a water supply and drainage connection relationship; Based on the direction of medium transport, the drainage network object corresponding to the drainage network object identifier is identified as the pollution source, the water supply network object corresponding to the water supply network object identifier is identified as the pollutant intrusion site, and the water supply network object located after the pollutant intrusion site along the direction of medium transport in the water supply event record sequence is identified as the affected water supply object, and the event record identifiers corresponding to the pollution source, the pollutant intrusion site and the affected water supply object are obtained respectively. Based on the isolation relationship, the event records that isolate the affected water supply objects are identified from the water supply event record sequence. The content of the event records is identified as the handling matters, and the pollution source, the part of the pollutant intrusion, the affected water supply object, the handling matters, and the corresponding event record identifiers are written into the event element generation instruction.

[0009] In one feasible implementation, after determining the operational topology state, including effective time periods, connectivity, isolation, and media transport direction, based on operational status records, the method further includes: For two adjacent operating topology states, change records are obtained by determining the connection status change of the same water supply and drainage connection relationship. The change records include the water supply and drainage connection relationship identifier, the status change time, and whether the water supply and drainage connection relationship changes from non-connected to connected or from connected to non-connected. Inputting the water supply event record sequence, drainage event record sequence, and event element generation instructions into the large language model generates cross-contamination event elements, including element categories, event element content, and corresponding event record identifiers, including: Input the water supply event record sequence, drainage event record sequence, change record, and event element generation instructions into the large language model to generate cross-contamination event elements that include element categories, event element content, and corresponding event record identifiers.

[0010] In one feasible implementation, the water supply event record sequence, drainage event record sequence, change records, and event element generation instructions are input into the large language model to generate cross-contamination event elements, including element categories, event element content, and corresponding event record identifiers, including: According to the order of the water supply event record sequence and the drainage event record sequence, the event record identifier, record time, pipeline object identifier and record content of each event record are combined into an event record item; According to the event record identifiers corresponding to the pollution source, pollutant intrusion site, affected water supply object and disposal item in the event element generation instruction, the target event record item is determined from the event record items, and the water supply and drainage connection relationship identifier and status change time in the change record are combined with the target event record item according to the medium transportation direction to form the event element generation content; Input the water supply event record sequence, drainage event record sequence, change record, event element generation instruction, and event element generation content into the large language model. The large language model then generates event element content and corresponding event record identifiers according to the element categories in the event element generation instruction, thus obtaining cross-contamination event elements.

[0011] In one feasible implementation, the water supply and drainage connection relationship identifier and status change time in the change record are combined with the target event record item according to the media transport direction to generate event element content, including: The change record from disconnected to connected is combined with the target event record items corresponding to the pollution source, the location of pollutant intrusion, and the affected water supply object to generate event element content; The change record that changed from connected to disconnected is combined with the target event record item corresponding to the handling item to generate event element content; Arrange the target event record items corresponding to the pollution source, the location of pollutant intrusion, and the affected water supply object in the event element generation content in the order of media transportation direction.

[0012] In one feasible implementation, cross-contamination event elements corresponding to the same water supply and drainage connection relationship are input into a large language model to generate a cross-contamination event summary, including: Based on the corresponding event record identifier in each cross-contamination event element, determine the recording time of the event record corresponding to each cross-contamination event element, and arrange each cross-contamination event element according to the recording time to obtain the event element arrangement content; Based on the element categories in the event element arrangement content, determine the event element content and corresponding event record identifiers corresponding to the pollution source, pollutant intrusion site, affected water supply object and treatment matter, and write the event element content and corresponding event record identifiers into the summary generation instruction in the order of pollution source, pollutant intrusion site, affected water supply object and treatment matter. The summary generation instructions and the arrangement of event elements are input into the large language model. The large language model then generates a summary of the cross-contamination event, including the pollution source, the location of the pollutant intrusion, the affected water supply objects, the handling matters, and the corresponding event record identifier, according to the summary generation instructions.

[0013] Secondly, this application provides an automatic event element summarization generation system based on a large language model, comprising: The acquisition module is used to acquire event records and operation status records of the water supply and drainage network within the target time period. The event records include event record identifier, recording time, network object identifier, and recording content. The determination module is used to determine the operating topology status, including the effective time period, connectivity, isolation, and media delivery direction, based on the operating status records. The sorting module is used to sort event records whose recording time falls within the valid time period into water supply event record sequence and drainage event record sequence for each operating topology state, according to the pipeline object identifier, connectivity relationship and medium delivery direction; The generation module is used to generate event element generation instructions based on the water supply and drainage connectivity relationship between the water supply event record sequence and the drainage event record sequence, according to the media transport direction and isolation relationship. The event element generation instructions include pollution source, pollutant intrusion site, affected water supply object, disposal matters, and their respective event record identifiers. The generation module is also used to input water supply event record sequences, drainage event record sequences and event element generation instructions into the large language model to generate cross-contamination event elements including element categories, event element content and corresponding event record identifiers; The generation module is also used to input cross-contamination event elements corresponding to the same water supply and drainage connection relationship into the large language model to generate cross-contamination event summaries.

[0014] Thirdly, this application provides an electronic device, comprising: Memory, used to store computer programs; A processor is used to implement the steps of the automatic event element summarization method based on a large language model as described in the first aspect above when executing a computer program.

[0015] The automatic event element summary generation method based on a large language model provided in this application is applicable to urban underground water supply and drainage network operation and maintenance and environmental monitoring scenarios. Abnormal drainage discharge monitoring, water supply quality monitoring, and valve operation records reflect different stages of pollution events. Valve opening and closing and isolation operations also change the connectivity between water supply and drainage network objects. This application determines the connectivity, isolation, and media transport direction within the effective time period based on the operation status records, and organizes water supply and drainage event records accordingly, associating abnormal environmental monitoring information with the network operation status of the corresponding time period. Furthermore, it inputs the pollution source, pollutant intrusion site, affected water supply objects, handling matters, and corresponding event record identifiers into the large language model, enabling the cross-contamination event summary to reflect the correlation between abnormal discharge, water quality alarms, affected objects, and handling activities. Therefore, this application improves the completeness of the association between environmental monitoring information and operation and maintenance event elements in water supply and drainage cross-contamination events and enhances the accuracy of the summary content.

[0016] Furthermore, this application identifies change records of water supply and drainage connectivity from disconnected to connected or vice versa by comparing adjacent operational topology states. These change records, along with water supply event record sequences, drainage event record sequences, and event element generation instructions, are input into a large language model. This allows the large language model to understand drainage anomaly discharge monitoring records, water supply quality monitoring alarm records, and isolation and treatment records in conjunction with the changes in water supply and drainage connectivity. Consequently, pollution sources, pollutant intrusion sites, affected water supply objects, and treatment matters can be correlated with the establishment or dissolution of connectivity, reducing the possibility of erroneous associations between environmental monitoring events and maintenance events under different operational states. Therefore, this application improves the temporal correlation completeness of water supply and drainage cross-contamination events and the accuracy of event summary content. Attached Figure Description

[0017] To more clearly illustrate the technical solutions of the embodiments of this application or the prior art, the accompanying drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the following drawings are only some embodiments of this application, and those skilled in the art can obtain other drawings based on these drawings without creative effort.

[0018] Figure 1 A flowchart illustrating an automatic event element summarization method based on a large language model, provided for embodiments of this application; Figure 2 A schematic diagram illustrating a specific implementation of an automatic event element summarization method based on a large language model, provided in this application embodiment; Figure 3 A schematic diagram illustrating a specific implementation of another method for automatically summarizing event elements based on a large language model, provided in this application embodiment; Figure 4 A schematic diagram of the structure of an automatic event element summarization system based on a large language model provided in this application embodiment; Figure 5 This is a schematic diagram of the hardware structure of an electronic device provided in one embodiment of this application. Detailed Implementation

[0019] To enable those skilled in the art to better understand the present application, the present application will be further described in detail below with reference to the accompanying drawings and specific embodiments. Obviously, the following embodiments are only a part of the embodiments of the present application, and not all of them. Other embodiments obtained by those skilled in the art based on the embodiments of the present application without creative effort are all within the protection scope of the present application.

[0020] To address the problems of existing technologies, embodiments of this application provide a method, system, and electronic device for automatically generating event element summaries based on a large language model. The method for automatically generating event element summaries based on a large language model, provided in this application, will be described first below.

[0021] Figure 1 This illustration shows a flowchart of an automatic event element summarization method based on a large language model, according to an embodiment of this application. Figure 1 As shown, the method includes steps S110 to S160.

[0022] In urban underground water supply and drainage network operation and maintenance and environmental monitoring scenarios, when abnormal discharge is detected by drainage outlet monitoring, water quality monitoring triggers alarms, or sewage overflow or connection errors occur in the drainage network, pollutants may enter the water supply network through the cross-connections of water supply and drainage systems. Valve opening, closing, and isolation operations alter the connection status between network components. Maintenance personnel need to combine environmental monitoring alarm records, inspection work orders, water quality monitoring records, drainage discharge monitoring records, and valve operation records to record the drainage components, water supply components, monitoring anomalies, and response activities involved in the pollution incident.

[0023] S110. Obtain event records and operation status records of the water supply and drainage network within the target time period. The event records include event record identifier, recording time, network object identifier, and recording content.

[0024] The target time period is a continuous time interval that defines the scope of data collection, determined by the start and end times. The start time of the target time period can be denoted as T0, and the end time as T1. A water supply and drainage network is a media transport network composed of water supply network objects, drainage network objects, valves, and connection nodes. Event logs record data on abnormal discharges, water quality anomalies, inspection findings, valve operations, or isolation and disposal activities. An event log identifier uniquely corresponds to each event log, and the log content records the event phenomenon, anomaly description, or handling procedures. When event logs originate from environmental monitoring, the log content also includes the monitoring point identifier corresponding to the network object identifier, monitoring item, measured value, measured value unit, alarm threshold, and alarm status.

[0025] The operation status record is data that describes the valve connection status and the medium delivery status. It may include the status record time, the pipeline object identifiers at both ends of the valve, the valve opening and closing status, the pipeline object identifier at the starting point of medium delivery, and the pipeline object identifier at the ending point of medium delivery.

[0026] The identification of pipeline network objects at both ends of the valve is matched with the valve identification, connection node identification, and pipe segment identification in the pipeline network facility ledger. When a water supply and drainage connection is formed between the drainage pipe segment and the water supply pipe segment, the operating status record is also linked to the inspection confirmation record of the cross connection or the pipeline network facility ledger record. The pipeline network object identification at the starting point and the ending point of the medium transportation is determined based on the flow monitoring record, pressure monitoring record, liquid level monitoring record, or on-site verification record on both sides of the cross connection when the valve is open; when the monitoring results show that the medium flows from the drainage side to the water supply side, the drainage pipeline network object identification is written into the medium transportation starting point pipeline network object identification, and the water supply pipeline network object identification is written into the medium transportation ending point pipeline network object identification.

[0027] The status recording time is used to indicate the time when the running status occurred or was collected, and it corresponds to different types of records, respectively, as the recording time in the event log. Multiple recording times in the event log can be recorded as Te1, Te2, Te3, and Te4 in sequence, while multiple status recording times in the running status log can be recorded as Tr1, Tr2, Tr3, and Tr4 in sequence. Te1 can correspond to a specific collection time of abnormal emission within the target time period, and Tr1 can correspond to the start time of the target time period.

[0028] During implementation, the target time period is first determined based on the event summary task, for example, T0 to T1. Next, event records are retrieved from alarm records, inspection work orders, water quality monitoring records, drainage discharge monitoring records, valve operation records, and pipeline operation records, retaining the event record identifier, recording time, pipeline object identifier, and record content. The monitoring point identifiers in the water quality monitoring records and drainage discharge monitoring records are written into the record content based on the correspondence between monitoring points and pipeline objects. The record content retains the monitoring item, detection value, detection value unit, alarm threshold, and alarm status.

[0029] Multiple event records constitute an event record set, which can be represented as E={E101, E102, E103, E104}. Simultaneously, operational status records are obtained from valve status records and media delivery records, retaining the pipeline object identifiers at both ends of the valve, the valve's open / closed status, and the start and end points of media delivery. Multiple operational status records constitute an operational status record set, which can be represented as R={R101, R102, R103, R104, R105, R106, R107, R108}. In this embodiment, event records are recorded in the order of Te1, Te2, Te3, and Te4, and operational status records are recorded in the order of Tr1, Tr2, Tr3, and Tr4. Te4 and Tr4 can represent event records and operational status records at the same time. For example, event records and operational status records can be shown in Tables 1 and 2.

[0030] Table 1 Event Log

[0031] Table 2 Example of running status record

[0032] The event records in Table 1 represent abnormal discharges, abnormal water quality, and isolation and treatment activities. Valve V1 in Table 2 is an isolation valve located at an abnormal cross-connection between drainage pipe section D1 and water supply pipe section S1. This abnormal cross-connection is confirmed through the pipeline network facility ledger and inspection records and is not considered a normal connection structure between the water supply and drainage networks. Valve V2 is located between water supply pipe sections S1 and S2. When valve V1 is open, the flow rate from D1 to S1 is determined by combining the flow monitoring records, pressure monitoring records, or on-site verification records on both sides of the cross-connection. D1 and S1 are then written into the pipeline network object identifiers for the starting and ending points of the medium transport, respectively. When the valve is closed, the starting and ending points of the medium transport only retain the potential transport paths configured in the pipeline network and do not indicate actual medium transport in the closed state.

[0033] S120. Based on the operating status record, determine the operating topology status, including the effective time period, connectivity, isolation, and media transport direction.

[0034] The operational topology status records the operational relationships of the water supply and drainage network according to time intervals, including the effective time period, connectivity, isolation, and media transport direction. The effective time period is a continuous time interval in which the valve opening and closing status does not change; connectivity indicates that media transport is possible between two network objects; isolation indicates that media transport is not possible between two network objects; and the media transport direction indicates the direction in which the media flows from the starting network object to the ending network object.

[0035] In one feasible implementation, step S120 includes: Based on the status record time in the operation status record, the pipeline network object identification at both ends of the valve, and the valve opening and closing status, determine the status change time when the valve opening and closing status changes, and divide the time period into time periods based on the start and end time of the target time period and the status change time. The state change time is the state record time corresponding to the later operating state record when the same valve has different valve opening and closing states in two adjacent operating state records. Time boundaries include the start time of the target time period, the end time of the target time period, and the state change time. A time period is a continuous time interval between two adjacent time boundaries, and the start time of the time period has a corresponding operating state record.

[0036] During implementation, the operating status records are first arranged according to valve identification and status recording time. For the same valve, the valve opening and closing status in adjacent operating status records are compared sequentially. Valve V1 is in the closed state at Tr1, in the open state at Tr2, and in the closed state at Tr3. Therefore, Tr2 and Tr3 are determined as the status change time of valve V1. Valve V2 is in the open state at Tr1, Tr2, and Tr3, and in the closed state at Tr4. Therefore, Tr4 is determined as the status change time of valve V2. The start time T0 of the target period, the status change times Tr2, Tr3, and Tr4, and the end time T1 of the target period are arranged chronologically to divide the time into four time periods: T0 to Tr2, Tr2 to Tr3, Tr3 to Tr4, and Tr4 to T1.

[0037] For each time period, based on the valve opening and closing status corresponding to the start time of the time period in the operation status record, for the pipeline objects at both ends of the valve in the open state, the connection relationship and the direction of medium transportation are determined according to the pipeline object identifiers of the starting point and the ending point of medium transportation; for the pipeline objects at both ends of the valve in the closed state, the isolation relationship is determined. The valve opening / closing status corresponding to the start time of the time period refers to the opening / closing status of each valve collected at the start time of the time period. The starting point and ending point pipeline object identifiers for medium transportation are used to determine the starting and ending objects of the medium transportation direction when the valve is open. When the valve is open, a connection is formed between the pipeline objects at both ends of the valve; when the valve is closed, an isolation is formed between the pipeline objects at both ends of the valve, and the starting point and ending point pipeline object identifiers for medium transportation are not used to determine the medium transportation direction when the valve is closed.

[0038] During implementation, for the time period from T0 to Tr2, the operating status record corresponding to Tr1 is read. Valve V1 is in the closed state, and the drainage pipe section D1 and the water supply pipe section S1 are isolated. Valve V2 is in the open state, and the water supply pipe section S1 and the water supply pipe section S2 are connected. The medium transportation direction is from the water supply pipe section S1 to the water supply pipe section S2. For the time period from Tr2 to Tr3, the operating status record corresponding to Tr2 is read. Both valves V1 and V2 are in the open state, and the drainage pipe section D1 and the water supply pipe section S1 are connected. The water supply pipe section S1 and the water supply pipe section S2 are also connected. The operating monitoring records on both sides of the cross connection indicate that the medium flows from the drainage pipe section D1 to the water supply pipe section S1, and the water supply network operating record indicates that the medium flows from the water supply pipe section S1 to the water supply pipe section S2. Therefore, the medium transportation direction is D1 to S1 and S1 to S2, respectively. For the time periods from Tr3 to Tr4 and from Tr4 to T1, the corresponding connectivity, isolation, and media transport direction are determined in the same manner.

[0039] Each time period is defined as a valid time period, and the valid time period, the corresponding connectivity, isolation, and media transport direction are defined as an operating topology state.

[0040] The connectivity, isolation, and media transport direction corresponding to the valid time period remain consistent within the valid time period. An operating topology state consists of a valid time period and the connectivity, isolation, and media transport direction within that valid time period. Different operating topology states are arranged in chronological order according to the valid time periods.

[0041] During implementation, T0 to Tr2 is defined as the first effective time period. The first operational topology includes the connectivity between water supply pipe segment S1 and water supply pipe segment S2, the isolation between drainage pipe segment D1 and water supply pipe segment S1, and the direction of medium transport from water supply pipe segment S1 to water supply pipe segment S2. Tr2 to Tr3 is defined as the second effective time period. The second operational topology includes the connectivity between drainage pipe segment D1 and water supply pipe segment S1, the connectivity between water supply pipe segment S1 and water supply pipe segment S2, and the direction of medium transport from drainage pipe segment D1 through water supply pipe segment S1 to water supply pipe segment S2. Tr3 to Tr4 and Tr4 to T1 are defined as the third and fourth effective time periods, respectively, and corresponding operational topologies are formed based on the valve opening / closing status at the start of each time period.

[0042] S130. For each operating topology state, according to the pipeline object identifier, connectivity relationship and medium delivery direction, the event records whose recording time is within the valid time period are arranged into a water supply event record sequence and a drainage event record sequence.

[0043] The water supply record set and drainage record set are record sets formed by dividing event records within the same valid time period according to the category of pipeline network objects. The arrangement of water supply pipeline network objects and drainage pipeline network objects is based on the order of pipeline network objects formed by connectivity and medium transportation direction. The water supply event record sequence and drainage event record sequence are ordered event record sets formed by arranging them based on the order of corresponding pipeline network objects.

[0044] In one feasible implementation, step S130 includes: Event records whose recording time falls within the valid time period are divided into a water supply record set and a drainage record set. The water supply record set includes event records of water supply network objects corresponding to network object identifiers, and the drainage record set includes event records of drainage network objects corresponding to network object identifiers.

[0045] The water supply record set and the drainage record set are two event record sets formed according to the type of pipeline network object. A record time within a valid time period means that the record time of an event is no earlier than the start time and earlier than the end time of the valid time period; for the last valid time period of a target period, the event record corresponding to the end time of the target time period can be included in the set.

[0046] During implementation, a valid time period corresponding to a running topology state is read, and event records whose recording times belong to the valid time period are filtered from the event logs. The network object type is identified based on the network object identifier. Water supply pipe segments with identifiers starting with "S" are assigned to the water supply record set, and drainage pipe segments with identifiers starting with "D" are assigned to the drainage record set. Taking the valid time period Tr2 to Tr3 as an example, the recording times of event records E101, E102, and E103 are Te1, Te2, and Te3, respectively, and Tr2 is earlier than Te1, Te2, and Te3, while Te1, Te2, and Te3 are earlier than Tr3. Therefore, E101, E102, and E103 are all filtered; E101 is assigned to the drainage record set, and E102 and E103 are assigned to the water supply record set. The resulting sets can be represented as the drainage record set {E101} and the water supply record set {E102, E103}.

[0047] According to the connectivity, the water supply network object identifiers in the water supply record set and the drainage network object identifiers in the drainage record set are arranged sequentially along the medium transport direction to obtain the water supply network object arrangement and the drainage network object arrangement. The arrangement of water supply network objects and drainage network objects is an ordered sequence of identifiers formed after determining the starting and ending network objects based on connectivity relationships. When multiple network objects are connected through continuous connectivity relationships, the ending object of the previous connectivity relationship is used as the starting object of the next connectivity relationship for sequential connection.

[0048] During implementation, the connectivity and media transport direction within the valid time period are read. Taking the valid time period Tr2 to Tr3 as an example, drainage pipe segment D1 is connected to water supply pipe segment S1, with the media transport direction from D1 to S1. Water supply pipe segment S1 is connected to water supply segment S2, with the media transport direction from S1 to S2. Therefore, the water supply network objects are arranged as {S1, S2}, and the drainage network objects are arranged as {D1}. When only the water supply pipe segment S1 and water supply segment S2 are connected within the valid time period Tr3 to Tr4, the water supply network objects are still arranged as {S1, S2}, and the drainage network objects are arranged as an empty sequence.

[0049] Arrange the event records in the water supply record set and the drainage record set according to the water supply network object arrangement and drainage network object arrangement respectively. Then, arrange the event records with the same network object identifier according to the recording time to obtain the water supply event record sequence and the drainage event record sequence.

[0050] The water supply event record sequence and the drainage event record sequence are ordered sequences formed by grouping and sorting event records according to the pipeline network object. First, the arrangement position between different pipeline network objects is determined according to the pipeline network object arrangement. Then, multiple event records corresponding to the same pipeline network object are arranged from earliest to latest according to the recording time. Each record item in the event record sequence still includes the event record identifier, recording time, pipeline network object identifier, and record content.

[0051] During implementation, event records in the water supply record set are arranged sequentially according to the water supply network object, and then multiple event records corresponding to the same water supply network object are sorted by recording time. Similarly, event records in the drainage record set are arranged sequentially according to the drainage network object, and then multiple event records corresponding to the same drainage network object are sorted by recording time. Taking the effective time period Tr2 to Tr3 as an example, the water supply record set includes E102 and E103, where E102 corresponds to S1 and E103 corresponds to S2. Therefore, the water supply event record sequence is represented as {E102, E103}. The drainage record set includes E101, so the drainage event record sequence is represented as {E101}. Taking the effective time period Tr4 to T1 as an example, event record E104 corresponds to water supply segment S2, so the water supply event record sequence is represented as {E104}, and the drainage event record sequence is represented as an empty sequence.

[0052] The water supply event record sequence and the drainage event record sequence are arranged separately according to each operating topology state, and event record sequences corresponding to different operating topology states are not directly merged. When generating event element generation instructions for the same water supply and drainage connection relationship, event records associated with the same water supply and drainage connection relationship are selected from the event record sequences corresponding to multiple operating topology states, and the valid time period corresponding to each event record is retained. Taking the water supply and drainage connection relationship formed by drainage pipe segment D1 and water supply pipe segment S1 as an example, E101, E102, and E103 come from the event record sequences corresponding to Tr2 to Tr3, and E104 comes from the water supply event record sequence corresponding to Tr4 to T1.

[0053] S140. For the water supply and drainage connection relationship between the water supply event record sequence and the drainage event record sequence, generate event element generation instructions according to the medium transport direction and isolation relationship. The event element generation instructions include pollution source, pollutant intrusion site, affected water supply object, disposal matters and their respective event record identifiers. Event element generation instructions are structured instructions based on the connectivity, isolation, and media transport direction of the same water supply and drainage connectivity relationship within the associated time range. These instructions specify the element categories, candidate pipeline objects, evidence record identifiers, and generation restrictions that the large language model needs to generate. The start time of the candidate associated time range is the time when the water supply and drainage connectivity relationship first changes from disconnected to connected; if the water supply and drainage connectivity relationship is already connected at the start time of the target time period, the start time of the candidate associated time range is the start time of the target time period.

[0054] The end time of the candidate association time range is the end time of the target time period. After determining the affected water supply objects and the isolation valves associated with the affected water supply objects based on the water supply and drainage connectivity, media transport direction, and isolation relationships within the candidate association time range, isolation event records are retrieved from the water supply event record sequence corresponding to each valid time period covered by the candidate association time range. The pipeline object identifier of the isolation event record is consistent with that of the affected water supply object, or the record content describes the isolation valve and the isolation operation. When an isolation event record is obtained, the recording time of the earliest recorded isolation event record is determined as the end time of the association time range; if no isolation event record is obtained and the water supply and drainage connectivity changes from connected to disconnected, the corresponding state change time is determined as the end time of the association time range; if the aforementioned state change time is not obtained, the end time of the target time period is determined as the end time of the association time range. The event records selected within the association time range maintain the corresponding operating topology status and valid time period information.

[0055] The water supply and drainage connectivity relationship indicates the connection between drainage network objects and water supply network objects that enables media transportation. The pollution source indicates a drainage network object experiencing abnormal emissions; the pollutant intrusion point indicates a water supply network object into which the drainage medium enters; the affected water supply object indicates a water supply network object located after the pollutant intrusion point along the media transportation direction; and the handling procedures indicate the isolation operations implemented for the affected water supply object. The event record identifiers corresponding to each event element are derived from event records associated with the network object or handling activities.

[0056] In one feasible implementation, step S140 includes: From the water supply event record sequence and the drainage event record sequence, identify the water supply network object identifier, the drainage network object identifier, and the corresponding event record identifier that have a water supply and drainage connection relationship; Water supply network object identifiers and drainage network object identifiers with water supply and drainage connectivity relationships refer to network object identifiers that appear in the water supply event record sequence and drainage event record sequence respectively, and are connected by the water supply and drainage connectivity relationship in the same operating topology state. The corresponding event record identifiers are those located in the same event record as the water supply network object identifier or drainage network object identifier.

[0057] During implementation, the water supply event record sequence and drainage event record sequence corresponding to the same operating topology state are read, and the water supply and drainage connectivity relationships are retrieved from the operating topology state. The drainage network object identifier in the water supply and drainage connectivity relationship is matched with the network object identifier in the drainage event record sequence, and the water supply network object identifier is matched with the network object identifier in the water supply event record sequence. Then, the event record identifiers in the matched event records are extracted. Taking the second operating topology state as an example, drainage pipe segment D1 and water supply pipe segment S1 form a water supply and drainage connectivity relationship. Drainage pipe segment D1 corresponds to event record E101, and water supply pipe segment S1 corresponds to event record E102. Therefore, the drainage network object identifier D1, the water supply network object identifier S1, and the event record identifiers E101 and E102 are obtained.

[0058] Based on the direction of medium transport, the drainage network object corresponding to the drainage network object identifier is identified as the pollution source, the water supply network object corresponding to the water supply network object identifier is identified as the pollutant intrusion site, and the water supply network object located after the pollutant intrusion site along the direction of medium transport in the water supply event record sequence is identified as the affected water supply object, and the event record identifiers corresponding to the pollution source, the pollutant intrusion site and the affected water supply object are obtained respectively. Pollution sources, contaminant intrusion sites, and affected water supply objects are event elements determined based on water supply and drainage connectivity and media transport direction. Pollution sources correspond to drainage network objects within the water supply and drainage connectivity, contaminant intrusion sites correspond to water supply network objects within the water supply connectivity, and affected water supply objects correspond to water supply network objects located after the contaminant intrusion site in the water supply event record sequence along the media transport direction. Each type of event element is associated with one or more event record identifiers.

[0059] During implementation, the drainage network object is first used as the starting point for media transport, and the water supply network object is searched along the media transport direction. In the second operating topology state, the media transport direction is from drainage segment D1 to water supply segment S1, and then from water supply segment S1 to water supply segment S2. Drainage segment D1 corresponds to event record E101. E101 records that the turbidity detection value of emission monitoring point M1 is 24 NTU and higher than the alarm threshold of 10 NTU. The pollution source is used as the element category to be generated, and drainage segment D1 is used as the candidate network object for the pollution source category. E101 is written into the evidence record identifier field corresponding to the pollution source category. Water supply segment S1 is the first water supply network object in the media transport direction from D1 to S1. E102 records that the turbidity detection value of water quality monitoring point M2 is 8 NTU and higher than the alarm threshold of 3 NTU. The pollutant intrusion site is used as the element category to be generated, and water supply segment S1 is used as the candidate network object for the pollutant intrusion site category. E102 is written into the evidence record identifier field corresponding to the pollutant intrusion site category. Water supply pipe section S2 is located after water supply pipe section S1. E103 records that the turbidity detection value of water quality monitoring point M3 is 7 NTU and higher than the alarm threshold of 3 NTU. The affected water supply object is taken as the element category to be generated, water supply pipe section S2 is taken as the candidate pipeline object of the affected water supply object category, and E103 is written into the evidence record identifier field corresponding to the affected water supply object category.

[0060] Based on the isolation relationship, the event records that isolate the affected water supply objects are identified from the water supply event record sequence. The content of the event records is identified as the handling matters, and the pollution source, the part of the pollutant intrusion, the affected water supply object, the handling matters, and the corresponding event record identifiers are written into the event element generation instruction.

[0061] The handling items are event record contents corresponding to the isolation relationship, used to describe operations such as closing valves, isolating water supply pipeline sections, or stopping media delivery. Event records for isolating affected water supply objects refer to event records where the pipeline object identifier is associated with the affected water supply object or isolation valve, and the record content describes the isolation operation. When the associated time range covers multiple valid time periods, the water supply event record sequence corresponding to each valid time period is read sequentially, and the isolation event record is retrieved from each water supply event record sequence. The event element generation instruction consists of the pollution source, the location of pollutant intrusion, the affected water supply object, the handling item, and the corresponding event record identifier.

[0062] During implementation, drainage pipe segment D1 and water supply pipe segment S1 change from disconnected to connected at Tr2. Based on this, Tr2 is determined as the start time of the candidate associated time range, and T1 is determined as the end time of the candidate associated time range. According to the media transport directions D1 to S1 and S1 to S2 in the second operating topology state, water supply pipe segment S2 is identified as the affected water supply object. Based on the isolation relationships in the corresponding operating topology states from Tr4 to T1, valve V2 is identified as the isolation valve for isolating water supply pipe segment S2. Then, the water supply event record sequences corresponding to Tr2 to Tr3, Tr3 to Tr4, and Tr4 to T1 are read sequentially to find the pipe network object identifier corresponding to water supply pipe segment S2, or the event record containing valve V2 and the isolation operation. E104 originates from the water supply event record sequence corresponding to Tr4 to T1, with a recording time of Te4, and the recording content states that valve V2 was closed and water supply pipe segment S2 was isolated. E104 is identified as the event record identifier corresponding to the disposal matter, and Te4 is identified as the end time of the associated time range, thus obtaining the associated time range from Tr2 to Te4. The disposal matter is taken as the element category to be generated, and the closure of valve V2 and isolation of water supply pipe segment S2 are taken as the candidate fact content of the disposal matter category. E104 is written into the evidence record identifier field corresponding to the disposal matter category.

[0063] S150. Input the water supply event record sequence, drainage event record sequence and event element generation instructions into the large language model to generate cross-contamination event elements including element categories, event element content and corresponding event record identifiers.

[0064] like Figure 2 As shown, the model input for generating cross-contamination event elements includes a water supply event record sequence, a drainage event record sequence, and an event element generation instruction. The model output is the cross-contamination event elements. For each water supply and drainage connection, the water supply event record sequence, drainage event record sequence, and event element generation instruction corresponding to the operating topology state covered by the final associated time range are read. Event records are written into the input text according to their respective order in the event record sequence, while retaining the valid time period to which the event record belongs. According to the order in the two types of event record sequences, the event record identifier, record time, pipeline object identifier, and record content of each event record are organized into event record items, while retaining the medium transport direction between the water supply pipeline object and the drainage pipeline object. After writing the event element generation instruction into the event record item, the input text for the large language model is formed. The large language model generates cross-contamination event elements including element category, event element content, and corresponding event record identifier based on the input text.

[0065] Large language models are text generation models that utilize large-scale text data for pre-training and record semantic relationships between words, phrases, event descriptions, and context through model parameters. Large language models can be pre-trained models that support Chinese text input, long text contexts, and instruction following, or language models adapted from texts related to water supply, drainage, and pipeline maintenance.

[0066] After receiving the input text, the large language model converts the event record items with field names, operation relationship items, and event element generation instructions into a sequence of lexical terms. Based on the monitoring point identifier, monitoring item, detection value, alarm status, inspection description, handling description, pipeline object identifier, and recording time in the event record items, it generates event element content corresponding to each element category. The pollution source, pollutant intrusion location, affected water supply object, and handling item in the event element generation instructions are used to limit the element category to be generated and the candidate pipeline object. The event record identifier is used to limit the scope of evidence records that can be referenced for each element category. The event element generation instructions are not pre-written into the event element content generated by the large language model.

[0067] In implementation, taking the water supply and drainage connection relationship formed by drainage pipe segment D1 and water supply pipe segment S1 as an example, the associated time range is Tr2 to Te4. The input text includes the event log section, the operational relationship section, the element generation requirements section, and the output format section in sequence. The event log section records the contents of E101, Te1, D1, and E101; the contents of E102, Te2, S1, and E102; the contents of E103, Te3, S2, and E103; and the contents of E104, Te4, S2, and E104. The operational relationship section records the water supply and drainage connection relationship formed by D1 and S1, the associated time range from Tr2 to Te4, the media transport direction from D1 to S1 and from S1 to S2, the state change from non-connected to connected corresponding to Tr2, and the state change from connected to non-connected corresponding to Tr3. The element generation requirements section includes four element categories: pollution source, pollutant intrusion site, affected water supply object, and treatment item, along with their corresponding E101, E102, E103, and E104 fields. The output format section includes three fields: element category, event element content, and corresponding event record identifier. The input text is then fed into the large language model.

[0068] The large language model generates cross-contamination event elements according to the element categories specified in the event element generation instructions. The generated results include event element content of four categories: pollution source, pollutant intrusion site, affected water supply object, and disposal matters. E101, E102, E103, and E104 are retained as corresponding event record identifiers, respectively.

[0069] The large language model outputs four cross-contamination event elements. The first cross-contamination event element is categorized as a pollution source, and its content is that abnormal turbidity emissions occur at monitoring point M1 corresponding to drainage pipe segment D1, with a turbidity detection value of 24 NTU. The corresponding event record is identified as E101. The second cross-contamination event element is categorized as a contaminant intrusion site, and its content is that water supply pipe segment S1 is located on the media transport path from D1 to S1, and abnormal turbidity occurs at its corresponding water quality monitoring point M2. The corresponding event record is identified as E102. The third cross-contamination event element is categorized as an affected water supply object, and its content is that water supply pipe segment S2 is located after water supply pipe segment S1, and abnormal turbidity occurs at its corresponding water quality monitoring point M3. The corresponding event record is identified as E103. The fourth cross-contamination event element is categorized as a handling matter, and its content is that valve V2 is closed and water supply pipe segment S2 is isolated. The corresponding event record is identified as E104.

[0070] In one feasible implementation, after step S120, the method further includes: S121. For two operating topology states that are adjacent in time, a change record is obtained by determining the change in the connectivity status of the same water supply and drainage connectivity relationship. The change record includes the water supply and drainage connectivity relationship identifier, the status change time, and whether the water supply and drainage connectivity relationship changes from non-connected to connected or from connected to non-connected.

[0071] Two adjacent operational topology states are those arranged sequentially according to their effective time periods and are consecutive. The same water supply and drainage connectivity relationship is a connection relationship consisting of the same drainage network object and the same water supply network object. Change records describe the state changes of the same water supply and drainage connectivity relationship between two adjacent operational topology states, including the water supply and drainage connectivity relationship identifier, the state change time, and the direction of the state change. The direction of the state change includes changing from disconnected to connected and from connected to disconnected.

[0072] During implementation, the operational topology status is read sequentially according to the effective time periods. The connectivity relationships in the previous and subsequent operational topology states are matched using the drainage network object identifier and the water supply network object identifier. For the same matched water supply / drainage connectivity relationship, the relationship status in the two operational topology states is compared. If the previous operational topology state is isolated and the subsequent operational topology state is connected, the start time of the subsequent effective time period is determined as the state change time from disconnected to connected. If the previous operational topology state is connected and the subsequent operational topology state is isolated, the start time of the subsequent effective time period is determined as the state change time from connected to disconnected. Multiple change records can be grouped into a change record set according to their state change times.

[0073] Taking the water supply and drainage connection relationship formed by drainage pipe segment D1 and water supply pipe segment S1 as an example, in the first operating topology state, D1 and S1 are isolated, while in the second operating topology state, D1 and S1 are connected, thus forming a first change record. The first change record includes the water supply and drainage connection relationship identifier formed by D1 and S1, the status recording time Tr2, and the direction of the state change from disconnected to connected. In the second operating topology state, D1 and S1 are connected, while in the third operating topology state, D1 and S1 are isolated, thus forming a second change record. The second change record includes the same water supply and drainage connection relationship identifier, the status recording time Tr3, and the direction of the state change from connected to disconnected. The set of change records can be represented as B = {B1, B2}.

[0074] Step S150 includes: S151, inputting the water supply event record sequence, drainage event record sequence, change record and event element generation instruction into the large language model to generate cross-contamination event elements including element category, event element content and corresponding event record identifier.

[0075] Figure 2 The model input-output relationship shown corresponds to the generation method of cross-contamination event elements without introducing change records. In the implementation corresponding to S151, in Figure 2 Based on the water supply event record sequence, drainage event record sequence, and event element generation instructions shown, change records are written into the input text of the large language model, while cross-contamination event elements are still output as models. In a further embodiment corresponding to S1513, after the event element generation content is formed, the event element generation content is written into the input text of the large language model.

[0076] S151 invokes the same large language model as S150. In the implementation corresponding to S151, the water supply event record sequence and drainage event record sequence provide environmental monitoring data, pipeline object identifiers, recording time, and recording content. The change record provides the time and direction of the state change for the establishment or termination of water supply and drainage connectivity. The event element generation instruction provides the category of element to be generated, candidate pipeline objects, and evidence record identifiers. The candidate pipeline objects and evidence record identifiers are used to limit the factual scope when the model is generated and are not pre-written into the event element content. The large language model generates the event element content corresponding to each element category based on the abnormal emission monitoring data, water quality monitoring data, isolation treatment descriptions, and connectivity state changes in the change records.

[0077] During implementation, the input text includes an event record sequence section, an operational relationship section, and an element generation requirement section. The event record sequence section sequentially writes event record items from the water supply event record sequence and the drainage event record sequence; the operational relationship section writes change records; and the element generation requirement section writes four element categories: pollution source, pollutant intrusion location, affected water supply object, and disposal item, as well as the candidate pipeline object and evidence record identifier corresponding to each element category. In a further implementation corresponding to S1513, the event element generation content formed in S1512 is written into the target event record section. The input content uses continuous text with field names to maintain the correspondence between the event record identifier, pipeline object identifier, record content, and status change information.

[0078] In one feasible implementation, step S151 includes: S1511. According to the order of arrangement in the water supply event record sequence and the drainage event record sequence, combine the event record identifier, record time, pipeline object identifier and record content of each event record into an event record item.

[0079] An event log entry is a record unit formed by combining the event log identifier, record time, pipeline object identifier, and record content from a single event record. Event log entries maintain the correspondence between the four types of fields to avoid separating the event record content from the source identifier. Water supply event log entries are arranged according to the sequence of water supply event records, and drainage event log entries are arranged according to the sequence of drainage event records.

[0080] During implementation, each event record in the water supply event record sequence and drainage event record sequence covered by the final associated time range is read sequentially, and four types of fields are extracted. Drainage event record E101 can be combined into an event record item including E101, Te1, D1, and an abnormal discharge description. Water supply event record E102 can be combined into an event record item including E102, Te2, S1, and a water quality anomaly description. Water supply event record E103 can be combined into an event record item including E103, Te3, S2, and a water quality anomaly alarm description. The handling event E104 originates from the water supply event record sequence corresponding to Tr4 to T1 and has been identified as an isolated event record during the associated time range determination process. It can be combined into an event record item including E104, Te4, S2, and the closure of valve V2 and isolation of water supply pipe segment S2. For the water supply and drainage connection relationship formed by drainage pipe segment D1 and water supply pipe segment S1, the water supply event record items and drainage event record items covered by the final associated time range Tr2 to Te4 are written into the same model input text; each event record item still retains the original arrangement order and corresponding valid time period in the event record sequence.

[0081] S1512. According to the event record identifiers corresponding to the pollution source, pollutant intrusion site, affected water supply object and disposal item in the event element generation instruction, determine the target event record item from the event record items, and combine the water supply and drainage connection relationship identifier and status change time in the change record with the target event record item according to the medium transportation direction to form the event element generation content.

[0082] A target event record is an event record whose event record identifier is the same as the event record identifier specified in the event element generation instruction. The event element generation content is structured input content for the large language model to read, consisting of target event record items, water supply and drainage connection relationship identifiers, state change time, state change direction, and media transport direction; among them, the target event record items corresponding to the pollution source, pollutant intrusion site, and affected water supply object are arranged according to the media transport direction, and the target event record items corresponding to the disposal items are de-associated according to the process of disconnecting the water supply and drainage connection relationship.

[0083] During implementation, the event record identifiers corresponding to the pollution source, pollutant intrusion site, affected water supply object, and treatment item in the event element generation instruction are first read. Then, the water supply event record items and drainage event record items are read sequentially, and the event record identifiers in the event record items are completely matched with the aforementioned event record identifiers. The matched event record items are determined as pollution source target event record items, pollutant intrusion site target event record items, affected water supply object target event record items, and treatment item target event record items, respectively. Taking pollution source corresponding to E101, pollutant intrusion site corresponding to E102, affected water supply object corresponding to E103, and treatment item corresponding to E104 as an example, the target event record items corresponding to E101, E102, E103, and E104 are determined respectively.

[0084] The change records from disconnected to connected are combined with the target event records of pollution sources, pollutant intrusion sites, and affected water supply objects, and arranged according to the media transport direction from drainage pipe section D1 to water supply pipe section S1 and from water supply pipe section S1 to water supply pipe section S2. The change records from connected to disconnected are combined with the target event records of the handling matters to form event element generation content, which includes the water supply and drainage connection relationship identifier, the time of status change, the direction of status change, and the event record content.

[0085] In one feasible implementation, step S1512 combines the water supply and drainage connection relationship identifier and status change time in the change record with the target event record item according to the media transport direction to generate event element content, including: The change record from disconnected to connected is combined with the target event record items corresponding to the pollution source, the location of pollutant intrusion, and the affected water supply object to generate event element content.

[0086] A change record from disconnected to connected indicates the establishment of a water supply and drainage connection. The target event record items corresponding to the pollution source, the location of pollutant intrusion, and the affected water supply object are determined based on the event record identifiers corresponding to the three types of elements in the event element generation instruction.

[0087] During implementation, the water supply and drainage connectivity identifier, Tr2, and the direction of the change from disconnected to connected status in B1 are read. Based on the event record identifier in the event element generation instruction, E101 corresponds to pollution source D1, E102 corresponds to pollutant intrusion site S1, and E103 corresponds to affected water supply object S2. The target event record items corresponding to B1, E101, E102, and E103 are combined and arranged in the direction from D1 to S1 and from S1 to S2 to form the event element generation content representing the abnormal discharge medium entering the water supply pipe section and being transported to the downstream water supply object.

[0088] The change record that changes from connected to disconnected and the target event record corresponding to the handling item are combined into event element generation content.

[0089] A change record from connected to disconnected indicates the termination of the water supply and drainage connection. The target event record item corresponding to the disposal item is determined based on the event record identifier corresponding to the disposal item in the event element generation instruction.

[0090] During implementation, the water supply and drainage connectivity identifier, status change time Tr3, and direction of change from connected to disconnected status are read from B2. Based on the water supply and drainage connectivity identifier, the contaminant intrusion site is determined to be water supply pipe segment S1, and based on the media transport direction D1 to S1 and S1 to S2, the affected water supply object is determined to be water supply pipe segment S2. E104 has already been identified as an isolation event record during the determination of the associated time range in S140, with a recording time of Te4 and a pipe network object identifier of S2. The recorded content includes the closure of valve V2 and the isolation of water supply pipe segment S2. The pipe network object identifier of E104 is consistent with the affected water supply object S2. Valve V2 is located between water supply pipe segments S1 and S2. The target event record items corresponding to the disposal matters in B2 and E104 are combined into event element generation content.

[0091] Arrange the target event record items corresponding to the pollution source, the location of pollutant intrusion, and the affected water supply object in the event element generation content in the order of media transportation direction.

[0092] Arranging by media transport direction means starting from the drainage network object where the pollution source is located, and sequentially arranging the target event record items corresponding to the pollutant intrusion site and the affected water supply object. The arrangement result includes the event record identifier, record time, network object identifier, and record content of each target event record item, and retains the status change time and status change direction corresponding to the change record.

[0093] During implementation, the target event record item corresponding to E101 is placed in position D1, the target event record item corresponding to E102 is placed in position S1, and the target event record item corresponding to E103 is placed in position S2, forming a sequence of target event record items. Tr2 in B1 and the direction of the change from disconnected to connected state are associated with the starting position of the target event record item sequence. Tr3 in B2 and the direction of the change from connected to disconnected state are associated with the handling item record item corresponding to E104, forming complete event element generation content.

[0094] S1513 inputs the water supply event record sequence, drainage event record sequence, change record, event element generation instruction, and event element generation content into the large language model. The large language model then generates the event element content and the corresponding event record identifier according to the element category in the event element generation instruction, thus obtaining the cross-contamination event element.

[0095] Cross-contamination event elements are structured outputs generated by the large language model according to the event element generation instructions. These elements include element category, event element content, and corresponding event record identifiers. The element category is derived from the pollution source, contaminant intrusion site, affected water supply objects, and handling procedures in the event element generation instructions. The event element content is a textual description generated by the large language model based on the event record content, pipeline object arrangement, media transport direction, and changes in connectivity. The corresponding event record identifier indicates the original event record upon which the generated content is based.

[0096] When generating cross-contamination event elements, the large language model first converts the input text into a sequence of terms, and then establishes semantic relationships based on field names, event record identifiers, pipeline object identifiers, and media delivery directions within the same input context. The event element generation instructions constrain the model's output, causing the model to generate content according to the specified four types of elements and referencing event record identifiers from the input records.

[0097] The model output uses a fixed field order: feature category, event feature content, and corresponding event record identifier. Feature categories are limited to pollution source, pollutant intrusion site, affected water supply object, and treatment item. The corresponding event record identifier must belong to the event record identifier set corresponding to that feature category in the event feature generation instruction. The pipeline object identifier, monitoring point identifier, monitoring item, detection value, detection value unit, treatment operation, and time information in the event feature content must originate from the input text. After the output content is generated, the feature category, corresponding event record identifier, and pipeline object identifier in the event record item are checked item by item; if there are undefined feature categories, missing event record identifiers, or missing pipeline object identifiers, the model is regenerated using the same input text and fixed field order.

[0098] In implementation, in the further implementation corresponding to S1513, the water supply event record sequence, drainage event record sequence, change record, event element generation instruction, and event element generation content are sequentially input into the large language model. The input content clearly defines E101 as the pollution source, E102 as the pollutant intrusion site, E103 as the affected water supply object, and E104 as the handling matter. It also provides the media transport direction from D1 to S1 and from S1 to S2, as well as the connection status change time for Tr2 and Tr3. The large language model generates pollution source event element content based on the record content of E101, pollutant intrusion site event element content based on the record content of E102, affected water supply object event element content based on the record content of E103, and handling matter event element content based on the record content of E104. E101, E102, E103, and E104 are retained as corresponding event record identifiers, ultimately resulting in a set of cross-contamination event elements.

[0099] S160. Input the cross-contamination event elements corresponding to the same water supply and drainage connection relationship into the large language model to generate a cross-contamination event summary.

[0100] like Figure 3 As shown, the model input for generating cross-contamination event summaries is cross-contamination event elements corresponding to the same water supply and drainage connectivity, and the model output is a cross-contamination event summary. Cross-contamination event elements corresponding to the same water supply and drainage connectivity are a set of event elements generated within the same associated time frame, based on the water supply and drainage connectivity between the same drainage network object and the same water supply network object. The cross-contamination event summary is an event description generated based on the pollution source, the location of pollutant intrusion, the affected water supply object, and the handling procedures within the same set. The event element arrangement is an ordered content formed by arranging the cross-contamination event elements according to the event recording time. The summary generation instruction is the input specifying the event element categories, content order, and event record identifier retention method in the cross-contamination event summary.

[0101] In one feasible implementation, step S160 includes: Based on the corresponding event record identifier in each cross-contamination event element, determine the recording time of the event record corresponding to each cross-contamination event element, and arrange each cross-contamination event element according to the recording time to obtain the event element arrangement content; The corresponding event record identifier is used to uniquely identify the event record that generated the cross-contamination event element. The event element arrangement includes the element category, event element content, corresponding event record identifier, and event record recording time for each cross-contamination event element. When a cross-contamination event element corresponds to multiple event record identifiers, the recording time corresponding to each of the multiple event record identifiers is determined separately; cross-contamination event elements with the same recording time are arranged according to the order in which the cross-contamination event elements were generated.

[0102] During implementation, the cross-contamination event elements corresponding to the same water supply and drainage connection relationship are read. The corresponding event record identifier is extracted from each cross-contamination event element, and event records with the same event record identifier are searched in the event record set to obtain the recording time. Taking E101 (pollution source), E102 (pollutant intrusion site), E103 (affected water supply object), and E104 (treatment item) as an example, the recording times Te1, Te2, Te3, and Te4 are determined respectively. Following the chronological order of Te1, Te2, Te3, and Te4, the cross-contamination event elements corresponding to the pollution source, pollutant intrusion site, affected water supply object, and treatment item are sequentially written into the event element arrangement content.

[0103] Based on the element categories in the event element arrangement content, determine the event element content and corresponding event record identifiers corresponding to the pollution source, pollutant intrusion site, affected water supply object and treatment matter, and write the event element content and corresponding event record identifiers into the summary generation instruction in the order of pollution source, pollutant intrusion site, affected water supply object and treatment matter. The summary generation instruction includes the output order of pollution source, contaminant intrusion site, affected water supply objects, and treatment matters; the corresponding event element content and event record identifier for each category; and the generation restriction of generating a summary only based on the arrangement of event elements while retaining the event record identifier. The pollution source, contaminant intrusion site, affected water supply objects, and treatment matters constitute the object transfer and treatment process of a cross-contamination event.

[0104] During implementation, the element categories in the event element arrangement are read, and the event element content and corresponding event record identifiers are extracted in the order of pollution source, pollutant intrusion site, affected water supply object, and treatment item. Pollution sources D1 and E101, pollutant intrusion sites S1 and E102, affected water supply objects S2 and E103, and treatment items closing valve V2 and isolating water supply pipe section S2 and E104 are written into the summary generation command. The summary generation command also stipulates that each type of event element should retain its corresponding event record identifier, and that network objects, treatment operations, or event record identifiers outside the event element arrangement content should not be written.

[0105] The summary generation instructions and the arrangement of event elements are input into the large language model. The large language model then generates a summary of the cross-contamination event, including the pollution source, the location of the pollutant intrusion, the affected water supply objects, the handling matters, and the corresponding event record identifier, according to the summary generation instructions.

[0106] The summary generation command and the event element arrangement content constitute the input text for the large language model. The summary generation command sequentially writes four summary fields: pollution source, pollutant intrusion site, affected water supply object, and treatment item, along with the corresponding event record identifier for each field. The event element arrangement content includes the monitoring point identifier, monitoring item, detection value, detection value unit, alarm status, pipeline object identifier, status change time, and treatment item for each event element. When the large language model generates the summary, the event record identifier in the summary fields must match the event record identifier in the event element arrangement content, and the monitoring data and pipeline object identifier in the summary text must originate from the event element arrangement content. The summary of a cross-contamination event includes the pollution source, pollutant intrusion site, affected water supply object, treatment item, and corresponding event record identifier.

[0107] Combination Figure 3 First, the event element arrangement content and summary generation instructions are generated from the cross-contamination event elements corresponding to the same water supply and drainage connection relationship. The summary generation instructions are written into the instruction part of the input text, and the event element arrangement content is written into the fact part of the input text. The input text is then input into the large language model. The large language model first identifies the order of occurrence of abnormal discharge, water quality abnormality alarm, and isolation treatment based on the event record time. Then, according to the summary generation instructions, it sequentially organizes the pollution source, the location of pollutant intrusion, the affected water supply object, and the treatment matters, and retains the corresponding event record identifier after the corresponding content.

[0108] In this embodiment, the cross-contamination event summary generated by the large language model can be as follows: the water supply and drainage connection between drainage pipe segment D1 and water supply pipe segment S1 changes from non-connected to connected at Tr2, and the medium transport direction is from D1 to S1. The turbidity detection value recorded at discharge monitoring point M1 corresponding to drainage pipe segment D1 is 24 NTU, which is higher than the alarm threshold of 10 NTU, and is therefore identified as a pollution source, with the corresponding event record identifier being E101; the turbidity detection value recorded at water quality monitoring point M2 corresponding to water supply pipe segment S1 is 8 NTU, which is higher than the alarm threshold of 3 NTU, and is therefore identified as a contaminant intrusion site, with the corresponding event record identifier being E102; the turbidity detection value recorded at water quality monitoring point M3 corresponding to water supply segment S2 is 7 NTU, which is higher than the alarm threshold of 3 NTU, and is therefore identified as an affected water supply object, with the corresponding event record identifier being E103; the water supply and drainage connection between D1 and S1 changes from connected to non-connected at Tr3, and the maintenance personnel close valve V2 and isolate water supply pipe segment S2 at Te4, which is considered a disposal matter, with the corresponding event record identifier being E104.

[0109] The cross-contamination event summary organizes abnormal emissions, abnormal water supply, affected objects, and isolation measures in the same text according to the event record time, and retains the event record identifier corresponding to each item, so that the summary content can correspond to the original records in the event record set.

[0110] In this embodiment, drainage discharge monitoring records, water supply quality monitoring alarm records, inspection records, and isolation treatment records are organized according to the connectivity, isolation, and media transport direction within different effective time periods. For cases where water supply and drainage connectivity is established or terminated, change records and event records are organized according to the media transport direction. Cross-contamination event elements and corresponding event record identifiers are input into a large language model to generate a cross-contamination event summary containing the pollution source, the location of pollutant intrusion, the affected water supply object, and the treatment measures.

[0111] Figure 4 This is a schematic diagram illustrating a specific implementation of an event element automatic summarization system based on a large language model, as provided in this application embodiment. (Refer to...) Figure 4 The system may include: The acquisition module 410 is used to acquire event records and operation status records of the water supply and drainage network within the target time period. The event records include event record identifier, recording time, network object identifier and recording content. The determination module 420 is used to determine the operating topology status, including the effective time period, connectivity, isolation and media transport direction, based on the operating status record. The arrangement module 430 is used to arrange event records whose recording time is within the valid time period into a water supply event record sequence and a drainage event record sequence for each operating topology state, according to the pipeline object identifier, connectivity relationship and medium delivery direction. The generation module 440 is used to generate event element generation instructions according to the medium transport direction and isolation relationship for the water supply and drainage connection relationship between the water supply event record sequence and the drainage event record sequence. The event element generation instructions include pollution source, pollutant intrusion site, affected water supply object, disposal matters and their respective event record identifiers. The generation module 440 is also used to input the water supply event record sequence, the drainage event record sequence and the event element generation instruction into the large language model to generate cross-contamination event elements including element categories, event element content and corresponding event record identifiers; The generation module 440 is also used to input cross-contamination event elements corresponding to the same water supply and drainage connection relationship into the large language model to generate cross-contamination event summaries.

[0112] The event element automatic summarization generation system based on a large language model in this application is used to implement the aforementioned event element automatic summarization generation method based on a large language model. Therefore, the specific implementation of the event element automatic summarization generation system based on a large language model can be found in the embodiment section of the event element automatic summarization generation method based on a large language model above. The specific implementation can be referred to the description of the corresponding embodiment, and will not be repeated here.

[0113] Figure 5 A schematic diagram of the hardware structure of an electronic device provided in one embodiment of this application is shown.

[0114] The electronic device may include a processor 510 and a memory 520 storing computer program instructions.

[0115] Specifically, the processor 510 may include a central processing unit (CPU), an application-specific integrated circuit (ASIC), or one or more integrated circuits that can be configured to implement the embodiments of this application.

[0116] Memory 520 may include mass storage for data or instructions. For example, and not as a limitation, memory 520 may include a hard disk drive (HDD), a floppy disk drive, flash memory, optical disk, magneto-optical disk, magnetic tape, or a Universal Serial Bus (USB) drive, or a combination of two or more of these.

[0117] Memory may include read-only memory (ROM), random access memory (RAM), disk storage media devices, optical storage media devices, flash memory devices, and electrical, optical, or other physical / tangible memory storage devices.

[0118] The processor 510 reads and executes computer program instructions stored in the memory 520 to implement any of the event element automatic summary generation methods based on a large language model in the above embodiments.

[0119] In one example, the electronic device may also include a communication interface 530 and a bus 540. Wherein, such as Figure 5 As shown, the processor 510, memory 520, and communication interface 530 are connected through bus 540 and complete communication with each other.

[0120] The communication interface 530 is mainly used to realize communication between various modules, devices, units and / or equipment in the embodiments of this application.

[0121] Bus 540 includes hardware, software, or both, which couples the components of an electronic device together.

[0122] The electronic device can execute the event element automatic summarization method based on a large language model in the embodiments of this application, thereby realizing the event element automatic summarization method based on a large language model as described in conjunction with the accompanying drawings.

[0123] It should be clarified that this application is not limited to the specific configurations and processes described above and shown in the figures. For the sake of brevity, detailed descriptions of known methods are omitted here. In the above embodiments, several specific steps are described and shown as examples. However, the method process of this application is not limited to the specific steps described and shown. Those skilled in the art can make various changes, modifications, and additions, or change the order of steps, after understanding the spirit of this application.

[0124] The functional blocks shown in the above-described block diagram can be implemented as hardware, software, firmware, or a combination thereof. When implemented in hardware, they can be, for example, electronic circuits, application-specific integrated circuits (ASICs), appropriate firmware, plug-ins, function cards, etc. When implemented in software, the elements of this application are programs or code segments used to perform the required tasks. The programs or code segments can be stored in a machine-readable medium or transmitted over a transmission medium or communication link via data signals carried on a carrier wave.

[0125] It should also be noted that the exemplary embodiments mentioned in this application describe methods or systems based on a series of steps or apparatus. However, this application is not limited to the order of the above steps; that is, the steps can be performed in the order mentioned in the embodiments, or in a different order, or several steps can be performed simultaneously.

[0126] The above provides a detailed description of the method, system, and electronic device for automatically generating event element summaries based on a large language model, as provided in this application. Specific examples have been used to illustrate the principles and implementation methods of this application. The descriptions of the embodiments above are merely for the purpose of helping to understand the method and its core ideas. It should be noted that those skilled in the art can make various improvements and modifications to this application without departing from its principles, and these improvements and modifications also fall within the protection scope of this application.

Claims

1. A method for automatically generating event element summaries based on a large language model, characterized in that, include: Acquire event records and operation status records of the water supply and drainage network within the target time period. The event records include event record identifier, recording time, network object identifier, and recording content. Based on the operational status record, determine the operational topology status, including the effective time period, connectivity, isolation, and media delivery direction; For each of the aforementioned operating topology states, the event records whose recording times fall within the valid time period are arranged into a water supply event record sequence and a drainage event record sequence according to the pipeline object identifier, connectivity relationship, and media delivery direction; For the water supply and drainage connectivity relationship between the water supply event record sequence and the drainage event record sequence, an event element generation instruction is generated according to the medium transport direction and isolation relationship. The event element generation instruction includes the pollution source, the pollutant intrusion site, the affected water supply object, the disposal matter, and the corresponding event record identifier. The water supply event record sequence, the drainage event record sequence, and the event element generation instruction are input into the large language model to generate cross-contamination event elements including element categories, event element content, and corresponding event record identifiers; Input the cross-contamination event elements corresponding to the same water supply and drainage connection relationship into the large language model to generate a cross-contamination event summary.

2. The method according to claim 1, characterized in that, Based on the operational status record, the operational topology status, including effective time periods, connectivity relationships, isolation relationships, and media transport direction, is determined, including: Based on the status record time, pipeline object identifiers at both ends of the valve, and valve opening / closing status in the operation status record, determine the status change time when the valve opening / closing status changes, and divide the time period into time periods based on the start and end time of the target time period and the status change time. For each time period, based on the valve opening / closing status corresponding to the start time of the time period in the operation status record, for the pipeline objects at both ends of the valve in the open state, the connection relationship and the direction of medium transportation are determined according to the pipeline object identifier of the starting point of medium transportation and the pipeline object identifier of the ending point of medium transportation; for the pipeline objects at both ends of the valve in the closed state, the isolation relationship is determined. Each time period is defined as a valid time period, and the valid time period, the corresponding connectivity, isolation, and media transport direction are defined as a running topology state.

3. The method according to claim 1, characterized in that, For each of the aforementioned operating topology states, the event records whose recording times fall within the valid time period are arranged into a water supply event record sequence and a drainage event record sequence according to the pipeline object identifier, connectivity, and media transport direction, including: The event records whose recording time is within the valid time period are divided into a water supply record set and a drainage record set. The water supply record set includes the event records of the water supply pipeline object corresponding to the pipeline object identifier, and the drainage record set includes the event records of the drainage pipeline object corresponding to the pipeline object identifier. According to the connectivity relationship, the water supply network object identifiers in the water supply record set and the drainage network object identifiers in the drainage record set are arranged sequentially along the medium transport direction to obtain the water supply network object arrangement and the drainage network object arrangement. According to the arrangement of the water supply network objects and the arrangement of the drainage network objects, the event records in the water supply record set and the drainage record set are arranged respectively, and the event records with the same network object identifier are arranged according to the record time to obtain the water supply event record sequence and the drainage event record sequence.

4. The method according to claim 1, characterized in that, Based on the water supply and drainage connectivity between the water supply event record sequence and the drainage event record sequence, an event element generation instruction is generated according to the medium transport direction and isolation relationship, including: From the water supply event record sequence and the drainage event record sequence, determine the water supply network object identifier, the drainage network object identifier, and the corresponding event record identifier that have the water supply and drainage connection relationship; Based on the medium transport direction, the drainage network object corresponding to the drainage network object identifier is identified as the pollution source, the water supply network object corresponding to the water supply network object identifier is identified as the pollutant intrusion site, and the water supply network object located after the pollutant intrusion site along the medium transport direction in the water supply event record sequence is identified as the affected water supply object, thereby obtaining the event record identifiers corresponding to the pollution source, the pollutant intrusion site, and the affected water supply object, respectively. Based on the isolation relationship, the event record for isolating the affected water supply object is determined from the water supply event record sequence, the record content of the event record is determined as the handling matter, and the pollution source, the pollutant intrusion site, the affected water supply object, the handling matter, and the corresponding event record identifier are written into the event element generation instruction.

5. The method according to claim 1, characterized in that, After determining the operating topology status, including the effective time period, connectivity, isolation, and media transport direction, based on the operating status record, the method further includes: For two adjacent operating topology states, a change record is obtained by determining the connection status change of the same water supply and drainage connection relationship. The change record includes the water supply and drainage connection relationship identifier, the status change time, and whether the water supply and drainage connection relationship changes from non-connected to connected or from connected to non-connected. The step involves inputting the water supply event record sequence, the drainage event record sequence, and the event element generation instruction into a large language model to generate cross-contamination event elements, including element categories, event element content, and corresponding event record identifiers. The water supply event record sequence, the drainage event record sequence, the change record, and the event element generation instruction are input into the large language model to generate cross-contamination event elements that include element categories, event element content, and corresponding event record identifiers.

6. The method according to claim 5, characterized in that, The water supply event record sequence, the drainage event record sequence, the change record, and the event element generation instruction are input into the large language model to generate cross-contamination event elements, including element categories, event element content, and corresponding event record identifiers, including: According to the arrangement order in the water supply event record sequence and the drainage event record sequence, the event record identifier, record time, pipeline object identifier and record content of each event record are combined into an event record item; According to the event record identifiers corresponding to the pollution source, pollutant intrusion site, affected water supply object and disposal item in the event element generation instruction, the target event record item is determined from the event record items, and the water supply and drainage connection relationship identifier and the status change time in the change record are combined with the target event record item according to the medium transport direction to form the event element generation content; The water supply event record sequence, the drainage event record sequence, the change record, the event element generation instruction, and the event element generation content are input into the large language model. The large language model generates event element content and corresponding event record identifiers according to the element categories in the event element generation instruction, thereby obtaining cross-contamination event elements.

7. The method according to claim 6, characterized in that, The event element generation content is generated by combining the water supply and drainage connection relationship identifier and the status change time in the change record with the target event record item according to the media transport direction, including: The change record that changes from disconnected to connected is combined with the target event record items corresponding to the pollution source, the pollutant intrusion site, and the affected water supply object to form the event element generation content; The change record that changes from connected to disconnected is combined with the target event record item corresponding to the disposal matter to generate the event element content; According to the media transport direction, the target event record items corresponding to the pollution source, the pollutant intrusion site, and the affected water supply object in the event element generation content are arranged sequentially.

8. The method according to claim 1, characterized in that, Inputting the cross-contamination event elements corresponding to the same water supply and drainage connection relationship into the large language model generates a cross-contamination event summary, including: Based on the corresponding event record identifier in each of the cross-contamination event elements, the recording time of the event record corresponding to each of the cross-contamination event elements is determined, and the cross-contamination event elements are arranged according to the recording time to obtain the event element arrangement content; Based on the element categories in the event element arrangement content, determine the event element content and corresponding event record identifier corresponding to the pollution source, pollutant intrusion site, affected water supply object and disposal matter, and write the event element content and the corresponding event record identifier into the summary generation instruction in the order of pollution source, pollutant intrusion site, affected water supply object and disposal matter; The summary generation instruction and the event element arrangement are input into the large language model, and the large language model generates a summary of the cross-contamination event, including the pollution source, the location of pollutant intrusion, the affected water supply object, the disposal matters, and the corresponding event record identifier, according to the summary generation instruction.

9. An automatic event element summarization system based on a large language model, characterized in that, include: The acquisition module is used to acquire event records and operation status records of the water supply and drainage network within a target time period. The event records include event record identifier, recording time, network object identifier, and recording content. The determination module is used to determine the operating topology status, including the effective time period, connectivity, isolation, and media delivery direction, based on the operating status record. The sorting module is used to sort the event records whose recording time is within the valid time period into a water supply event record sequence and a drainage event record sequence for each of the operating topology states, according to the pipeline object identifier, connectivity relationship and medium transport direction; The generation module is used to generate event element generation instructions based on the water supply and drainage connectivity relationship between the water supply event record sequence and the drainage event record sequence, according to the medium transport direction and isolation relationship. The event element generation instructions include pollution source, pollutant intrusion site, affected water supply object, disposal matters, and their respective event record identifiers. The generation module is also used to input the water supply event record sequence, the drainage event record sequence and the event element generation instruction into the big language model to generate cross-contamination event elements including element categories, event element content and corresponding event record identifiers; The generation module is also used to input the cross-contamination event elements corresponding to the same water supply and drainage connection relationship into the large language model to generate a cross-contamination event summary.

10. An electronic device, characterized in that, include: Memory, used to store computer programs; A processor, configured to execute the computer program to implement the steps of the automatic event element summarization method based on a large language model as described in any one of claims 1 to 8.

Images