An alarm management method based on a city internet of things platform

Abstract

The application discloses an alarm management method based on a city Internet of Things platform and particularly relates to the technical field of Internet of Things data processing, wherein by accessing multi-source sensors and devices in city infrastructure, real-time operation data is acquired, and the real-time operation data is subjected to abnormality detection to generate an initial alarm information set; alarm risk levels are calculated according to abnormal amplitude, space-time concentration degree and historical event frequency, and alarm task queues are formed in combination with alarm geographical location correlation degrees; further, a dispatch table is dynamically updated according to the alarm task queues to generate corresponding operation and maintenance instruction sets, and alarm events are uniformly dispatched and assigned; meanwhile, processing results and alarm data are fed back to the platform to iteratively update alarm priorities; the application can realize unified aggregation, intelligent sorting and dynamic dispatching of multi-source alarm information, reduce repeated order dispatching and response delay, and improve city infrastructure operation and maintenance efficiency.

Description

Technical Field

[0001] This invention relates to the field of Internet of Things (IoT) data processing technology, and more specifically to an alarm management method based on an urban IoT platform. Background Technology

[0002] With the intelligent development of urban infrastructure, urban IoT platforms have been widely used in scenarios such as traffic management, smart lighting, water supply and drainage, and environmental monitoring.

[0003] However, current urban facility alarm systems generally suffer from problems such as fragmented data, delayed response, and reliance on experience for manual processing. For example, municipal drainage networks may generate multiple simultaneous alarms during sudden rainstorms, but traditional systems cannot effectively aggregate and prioritize alarm events, leading to duplicate dispatches, delayed responses, and even missed critical maintenance opportunities, seriously threatening urban operational safety and the continuity of public services. Summary of the Invention

[0004] The purpose of this invention is to provide an alarm management method based on an urban Internet of Things (IoT) platform to address the shortcomings in the prior art.

[0005] To achieve the above objectives, the present invention provides the following technical solution: an alarm management method based on an urban Internet of Things (IoT) platform, comprising: Access multiple infrastructure sensors and devices in the city's Internet of Things (IoT) platform to obtain real-time operational data from each device; Anomaly detection is performed on the acquired real-time data to generate an initial set of alarm information, and the risk level and geographical location correlation of each alarm are calculated. The initial alarm information set is used to construct an alarm priority model based on risk level, correlation, and historical processing records, forming a schedulable alarm task queue. Based on the alarm task queue, the scheduling model corresponding to the city's IoT platform is dynamically updated, and a corresponding set of operation and maintenance instructions is generated. Based on the set of operation and maintenance instructions, all alarm events are uniformly scheduled and assigned, and the processing results and alarm data are fed back to the platform to iteratively update the alarm priority.

[0006] Preferably, anomaly detection is performed on the acquired real-time operational data, including the following steps: The real-time operating data of the same sensor within the same time slice are sorted according to the acquisition time, and data marked as missing or with communication delay are removed to obtain a valid numerical sequence. The short-term amplitude range is calculated based on the maximum and minimum monitoring values ​​within the effective numerical sequence, and the continuous stable duration is accumulated based on the changes in adjacent effective values. Based on the combination relationship between short-term amplitude range and continuous stable duration, abnormal information is classified into sudden abnormalities or slow drift abnormalities, thus obtaining abnormal information with abnormality type.

[0007] Preferably, generating an initial alarm information set includes the following steps: calculating the temporal overlap and spatial proximity of abnormal information within the same spatial area, and obtaining the spatiotemporal overlap value by multiplying the temporal overlap and spatial proximity; taking the earliest generated abnormal information as the starting point, grouping abnormal information with a spatiotemporal overlap value of 0.6 into the same abnormal event; and forming preliminary alarm entries based on the abnormal type, center location, earliest acquisition time, and list of involved sensors within the same abnormal event.

[0008] Preferably, the risk level of each alarm is calculated, including the following steps: The abnormal amplitude is calculated based on the absolute value of the difference between the actual monitored value and the normal reference value and the allowable deviation value. The normal reference value is the median of the effective monitored values ​​of the same equipment within the same time period. Based on the number of sensors involved in the abnormal event, the duration of the abnormal event, and the frequency of similar historical events, the abnormal amplitude conversion score, the spatiotemporal concentration conversion score, and the historical event frequency conversion score are obtained respectively, and the risk level is determined according to the risk score.

[0009] Preferably, the calculation of the geographic location correlation for each alarm includes the following steps: The alarm entries in the initial alarm information set are compared according to geographical coordinates and timestamps, and the alarm entries whose spatial distance does not exceed the registered service radius of the relevant facilities and whose alarm time difference does not exceed the specified association duration are selected as the neighboring alarm points; The density and service overlap ratio are obtained based on the number of nearby alarm points, the service area of ​​related facilities, and the overlapping service area. Based on the distance from the alarm entry to the center of the event cluster, the density, and the service overlap ratio, calculate the geographical location correlation and attach it to the corresponding alarm entry.

[0010] Preferably, forming a schedulable alarm task queue includes the following steps: The risk level is converted into a risk level conversion value, and a comprehensive score is calculated by combining the geographical location correlation and historical processing record scores. The historical processing record scores are obtained based on the average response time and number of times the same anomaly type is processed in the same spatial area. All alarm items are sorted from high to low according to the comprehensive score, and if the comprehensive scores are the same, they are sorted from early to late according to the alarm generation time to obtain the processing order of alarm tasks.

[0011] Preferably, forming a schedulable alarm task queue further includes the following steps: For alarm entries in the same spatial area whose comprehensive score reaches the merging reference value, calculate the spatial proximity value and the temporal proximity value; calculate the merging judgment value according to the spatial proximity value accounting for 60% and the temporal proximity value accounting for 40%, and merge alarm entries with a merging judgment value of 0.7 into a unified alarm task; take the alarm entry with the highest comprehensive score as the main alarm entry, take the remaining alarm entries as associated alarm entries, and write the unified alarm task into the alarm task queue.

[0012] Preferably, generating the corresponding set of operation and maintenance instructions includes the following steps: Read the first alarm task in the alarm task queue, and determine no more than 3 related facilities based on the distance between the alarm location and the facility's registered coordinates; calculate the operation score based on the risk level conversion value, geographical location correlation, task generation time sequence conversion value, facility distance conversion value, and facility status conversion value; determine the operation priority order of related facilities and sensors involved according to the operation score from high to low, and update the preprocessing status mark of the corresponding facility in the scheduling table to "pre-occupied".

[0013] Preferably, generating the corresponding set of operation and maintenance instructions also includes the following steps: Based on the current occupancy status and risk level of relevant facilities, generate operation and maintenance instructions to start equipment, stop equipment, adjust parameters, call backup facilities, or send on-site inspection requests; the parameter adjustment amount is calculated according to the equipment's registered rated parameters, abnormal amplitude conversion score, and 20% adjustment ratio, and the adjusted target parameters do not exceed the equipment's registered safe operation limit; bind the set of operation and maintenance instructions to alarm task entries, and write them into the sequence of instructions to be executed according to the order of the alarm task queue.

[0014] Preferably, the iterative update of alarm priorities includes the following steps: The system sequentially reads the maintenance instructions from the sequence of instructions to be executed, records the instruction issuance time, execution start time, execution end time, operation completion status, and any abnormal situations, and forms processing result data. The processing result data is then merged with the original data of the corresponding alarm events according to the task number to obtain a processing history record containing response time, processing time, completion status, and impact range correction values. Based on the updated processing history record, risk level, and geographical location correlation, the comprehensive score is recalculated, and incomplete and transferred alarm tasks are reordered, while completed alarm tasks are removed from the alarm task queue.

[0015] The technical effects and advantages provided by the present invention in the above technical solution are as follows: The core technical approach of this invention lies in performing refined anomaly detection on real-time operational data from multi-source urban IoT devices. It combines this with multi-dimensional features such as short-term amplitude variations within time slices, continuous stable duration, spatiotemporal overlap analysis, historical processing records, and geographic location correlation to construct alarm priorities and scheduling strategies. By quantifying anomaly amplitude, spatiotemporal concentration, and historical event frequency into risk levels, and generating a comprehensive score together with geographic location correlation, and then merging alarms based on spatial and temporal proximity to form a schedulable alarm task queue, this invention overcomes the shortcomings of existing technologies, such as isolated alarm information, delayed response, and reliance on manual experience, achieving unified aggregation and prioritized processing of multi-source alarm information.

[0016] Furthermore, this invention achieves dynamic scheduling and unified assignment of alarm events by binding a set of operation and maintenance instructions with alarm tasks. It also iteratively updates alarm priorities through feedback processing results, ensuring that high-risk, continuous, or recurring events receive priority responses. This technical approach directly solves the problems of duplicate task assignment, resource conflicts, and response delays in existing urban infrastructure operation and maintenance. It enhances the intelligent scheduling capabilities of urban IoT platforms for complex, multi-point emergencies, significantly improving alarm response efficiency, resource utilization, and processing accuracy, thus ensuring urban operational safety and the continuity of public services. Attached Figure Description

[0017] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this invention. For those skilled in the art, other drawings can be obtained based on these drawings.

[0018] Figure 1 This is a flowchart of an alarm management method based on a city Internet of Things platform according to the present invention. Detailed Implementation

[0019] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0020] Example 1, please refer to Figure 1 As shown in this embodiment, an alarm management method based on a city IoT platform includes: It connects to multiple infrastructure sensors and devices in the city's Internet of Things (IoT) platform to obtain real-time operating data from each device.

[0021] In this invention, the urban IoT platform pre-establishes a device access directory, which includes at least a unique device identifier, device type, installation location, affiliated facility, communication protocol, data acquisition frequency, and management responsibility unit. The platform connects to multi-source sensors and devices in the urban infrastructure via IoT gateways. These multi-source sensors and devices include, but are not limited to, drainage network level sensors, rain gauges, manhole cover status sensors, pump station operation controllers, road surface water monitoring equipment, road video analysis equipment, street light controllers, and power supply status detection equipment.

[0022] Specifically, taking the scenario of drainage in low-lying urban roads as an example, when short-term heavy rainfall occurs, rain gauges, manhole level sensors, drainage pump station controllers, and road surface water detection equipment in low-lying sections upload data such as rainfall, manhole level height, pump start / stop status, pump current, and road surface water depth to the urban IoT platform. Because different devices use different communication protocols and data formats, the platform uses an IoT gateway to parse and convert the received data, uniformly converting the data uploaded by each device into a standard operating data packet.

[0023] The standard operating data package includes a unique device identifier, device type, acquisition time, geographical location, real-time monitoring value, device operating status, communication status, and data quality identifier. The unique device identifier distinguishes different infrastructure devices; the geographical location is used to determine the spatial correlation between alarms; the acquisition time is used for time alignment of multi-source data; the real-time monitoring value is used for subsequent anomaly detection; and the data quality identifier determines whether the data is missing, delayed, or exhibits abnormal fluctuations.

[0024] Furthermore, the platform synchronously processes device data from different sampling frequencies according to a preset time window. For example, rainfall, liquid level, pump station status, and water depth values ​​uploaded within one minute in the same low-lying road section are grouped into a single operational data set for the same time period, resulting in a real-time operational data set to be analyzed. This real-time operational data set serves as the foundation for subsequent anomaly detection and alarm generation, enabling subsequent steps to make comprehensive judgments based on the status of multiple sources of equipment within the same spatial area and time window, avoiding false alarms based solely on momentary anomalies of a single device.

[0025] Anomaly detection is performed on the acquired real-time data to generate an initial set of alarm information, and the risk level and geographical location correlation of each alarm are calculated.

[0026] When performing anomaly detection on the acquired real-time data, the city's IoT platform uses preset time slices as processing units, arranging the real-time operational data uploaded within the same spatial area according to the order of acquisition time. Each time slice can be 60 seconds long, and adjacent time slices do not overlap. For sensor value sequences formed by the same sensor within the same time slice, data marked as missing or with communication delays are first removed, and then valid values ​​are retained from earliest to latest acquisition time. For the first valid value to the last valid value, the maximum and minimum monitoring values ​​within the slice are obtained respectively, and the short-term amplitude range is calculated by subtracting the minimum monitoring value within the slice from the maximum monitoring value.

[0027] The duration of continuous stability is determined based on the change between two adjacent valid values. The adjacent change is the absolute value of the subsequent valid value minus the previous valid value. When the adjacent change does not exceed the preset steady-state limit of the corresponding data item, the adjacent sampling interval is included in the stable duration; when the adjacent change exceeds the preset steady-state limit of the corresponding data item, the previous stable duration ends, and accumulation restarts. The longest stable duration obtained within a time slice is taken as the continuous stable duration. Anomalies are determined by combining the short-term amplitude range and the continuous stable duration: when the short-term amplitude range reaches the preset allowable amplitude limit of the corresponding data item, and the continuous stable duration is less than 30% of the time slice duration, it is marked as a sudden anomaly; when the average values ​​of three consecutive time slices change in the same direction, the difference between the first and last average values ​​of the three time slices reaches the preset allowable offset limit of the corresponding data item, and the continuous stable duration within each time slice is not less than 70% of the time slice duration, it is marked as a slow drift anomaly. The above processing can distinguish between instantaneous and drastic changes and continuous and slow deviations, avoiding the direct use of a single communication jitter as the basis for an alarm.

[0028] Upon receiving information about sudden or slow-drift anomalies, a spatiotemporal overlap analysis is performed on the anomaly information from different sensors within the same spatial area. The same spatial area can be determined based on low-lying road areas, drainage zones, pumping station service areas, or other registered facility management areas. Each anomaly information entry must include at least the device's unique identifier, anomaly type, acquisition time, geographical location, and anomaly data item. Temporal overlap is calculated by dividing the time intersection duration of two anomaly information entries by the time union duration of the two anomaly information entries; spatial proximity is calculated by subtracting the ratio of the distance between two anomaly locations to the maximum management distance of the corresponding spatial area from 1, with a value of 0 when the result is less than 0. The spatiotemporal overlap value of two anomaly information entries is calculated by multiplying the temporal overlap by the spatial proximity.

[0029] Starting with the earliest generated anomaly, anomalies located in the same spatial region and with a spatiotemporal overlap value of 0.6 are grouped into the same anomaly event. Newly grouped anomalies continue to undergo the same comparison until no further anomalies can be grouped. Each anomaly event forms one preliminary alarm entry. The anomaly type in the preliminary alarm entry is the anomaly type that occurs most frequently within the anomaly event; the location is the center position of the geographic coordinates of all sensors involved in the anomaly event; the initial time is the collection time of the earliest anomaly information within the anomaly event; the list of involved sensors is recorded according to the device's unique identifier and device type. The concentration level is calculated by dividing the number of sensors involved in the anomaly event by the number of sensors already connected in the same spatial region. The impact range is determined by the area of ​​the smallest external region formed by the sensor locations involved in the anomaly event; when the number of involved sensors is 1, the impact range is the registered service area area of ​​the facility corresponding to that sensor.

[0030] The abnormal amplitude of each preliminary alarm entry is calculated based on the abnormal information involved in the abnormal event. The abnormal amplitude of a single abnormal information is "the absolute value of the difference between the actual monitored value and the normal baseline value divided by the allowable deviation value". The normal baseline value is the median of the effective monitored values ​​of the same equipment in the same time period over the past 30 days. The allowable deviation value is the operational allowable deviation registered for this data item. The abnormal amplitude of the preliminary alarm entry is the average of the abnormal amplitudes involved in the abnormal event. The spatiotemporal concentration is calculated by "concentration degree multiplied by time concentration degree". The time concentration degree is "1 minus the ratio of the duration of the abnormal event to the specified response time of the same spatial area". If the result is less than 0, it is taken as 0. The historical event frequency is the number of times the same spatial area and the same abnormal type have formed preliminary alarm entries in the past 30 days.

[0031] The risk score is calculated as follows: The abnormal amplitude conversion score is calculated by multiplying the abnormal amplitude by 100 (or rounding down to 100 if the value exceeds 100); the spatiotemporal concentration conversion score is calculated by multiplying the spatiotemporal concentration score by 100; and the historical event frequency conversion score is calculated by multiplying the historical event frequency by 10 (or rounding down to 100 if the value exceeds 100). The risk score is equal to the abnormal amplitude conversion score multiplied by 45%, the spatiotemporal concentration conversion score multiplied by 35%, and the historical event frequency conversion score multiplied by 20%, then summed. The risk level is determined according to the score range: 0 to 20 points correspond to Level 1, over 20 to 40 points to Level 2, over 40 to 60 points to Level 3, over 60 to 80 points to Level 4, and over 80 to 100 points to Level 5. The risk level is then appended to the corresponding initial alarm entry to form an initial alarm information set containing the risk level.

[0032] When calculating the geographic correlation of each alarm, each alarm entry in the initial alarm information set is first bound to the geographic coordinates of the corresponding sensor, and the timestamp of the alarm is retained. The geographic coordinates are then converted to planar coordinates before being used in distance calculations. The spatial distance between two alarms is obtained by taking the square root of the sum of the squares of the horizontal and vertical coordinate differences. For any alarm entry within the same spatial area, other alarm entries whose spatial distance does not exceed the registered service radius of the relevant facility and whose alarm time difference does not exceed the specified correlation duration are selected as neighboring alarm points. The density is calculated by dividing the number of neighboring alarm points by the service area of ​​the relevant facility.

[0033] Event clusters are established from earliest to latest timestamp. First, the earliest alarm entry not yet assigned to an event cluster is selected as the starting entry. Then, neighboring alarm points of this starting entry are assigned to the same event cluster. Alarm entries assigned to clusters continue searching for their neighboring alarm points until no new alarm entries can be assigned. After each alarm entry is assigned to an event cluster, the service overlap ratio is calculated. The service overlap ratio is the sum of the overlapping areas of the service areas of the relevant facilities corresponding to the alarm entry and the service areas of the relevant facilities corresponding to other alarm entries within the same event cluster, divided by the sum of the service areas of the relevant facilities corresponding to the alarm entry and the service areas of the relevant facilities corresponding to other alarm entries within the same event cluster.

[0034] The proximity value for each alarm is calculated by subtracting the distance between the alarm entry and the center of the same event cluster from 1, divided by the registered service radius of the relevant facility. A value less than 0 is taken as 0. The center of the event cluster is the average coordinate of all alarm entries within the cluster. The density conversion value is calculated by dividing the density by the maximum density of all alarm entries within the same spatial area. When all density values ​​within the same spatial area are 0, the density conversion value is 0. The geographic location correlation is calculated by multiplying the proximity value by 50%, the density conversion value by 30%, and the service overlap ratio by 20%, then summing these values ​​and multiplying by 100. The value ranges from 0 to 100. The geographic location correlation is appended to the corresponding alarm entry to form an initial alarm information set containing risk level and geographic location correlation. This set is used for subsequent alarm priority sorting and scheduling, enabling the same or consecutive events to be merged and identified, reducing duplicate processing caused by continuous reporting from adjacent devices.

[0035] The initial set of alarm information is used to construct an alarm priority model based on risk level, correlation, and historical processing records, forming a schedulable alarm task queue.

[0036] In this embodiment, the alarm priority model consists of four processes: comprehensive score calculation, alarm item sorting, merging of similar alarms, and queue writing, which are used to convert the initial alarm information set into a schedulable alarm task queue.

[0037] A comprehensive score is calculated for each alarm entry in the initial alarm information set. The risk level is first converted to a risk level conversion value: Risk Level Conversion Value = Risk Level × 20. Therefore, for risk levels 1 to 5, the corresponding risk level conversion values ​​are 20, 40, 60, 80, and 100 points, respectively. Geographical location correlation uses the 0 to 100 point values ​​obtained in the previous step. Historical processing record scores are obtained based on processing records of the same spatial area and the same anomaly type within the past 30 days; when no similar alarms exist within the past 30 days, the historical processing record score is 50 points. When similar alarms exist, the average response time is first calculated: Average Response Time = Sum of the time between the first reception and processing time of each historical alarm and the alarm occurrence time ÷ Number of historical alarms; then, the response time score is calculated: Response Time Score = Average Response Time ÷ Specified Response Time × 100, with scores exceeding 100 points rounded up to 100 points. The specified response time is pre-registered according to facility type: 30 minutes for drainage alarms, 20 minutes for power supply alarms, and 15 minutes for manhole cover status alarms. Then, a completion gap score is calculated: Completion gap score = [1 - Number of completed responses ÷ Number of historical alarms] × 100. Historical processing record score = Response time score × 60% + Completion gap score × 40%. Overall score = Risk level conversion value × 50% + Geographical location correlation × 30% + Historical processing record score × 20%. The calculation result is rounded to two decimal places.

[0038] All alarm entries are sorted from highest to lowest based on their comprehensive score, forming an initial processing order. Alarm entries with the same comprehensive score are sorted from earliest to latest alarm occurrence time; alarm entries with the same comprehensive score and the same alarm occurrence time are sorted from highest to lowest risk level; if the risk levels are still the same, they are sorted from highest to lowest geographical location correlation. This sorting method prioritizes high-risk alarm entries, those with obvious spatial clustering, and those with historically low processing efficiency for subsequent scheduling.

[0039] Multiple alarm entries with a comprehensive score higher than the merged reference value within the same spatial area are merged. The preferred merged reference value is 75 points. When the number of unprocessed alarm entries in the same spatial area reaches 5 on a given day, the merged reference value is the average comprehensive score of the unprocessed alarm entries in that spatial area plus 10 points. If the calculated result is lower than 70 points, it is counted as 70 points; if it exceeds 85 points, it is counted as 85 points. For alarm entries with a comprehensive score reaching the merged reference value, the spatial distance and time difference between any two alarm entries are calculated. Spatial proximity value = 1 - spatial distance ÷ merged distance, where the merged distance is 60% of the registered service radius of the relevant facility; a spatial proximity value less than 0 is used as 0. Time proximity value = 1 - time difference ÷ 10; a time proximity value less than 0 is used as 0. Merging judgment value = spatial proximity value × 60% + time proximity value × 40%. Alarm entries with a merged judgment value of 0.7 or higher are merged into the same unified alarm task. The unified alarm task retains the alarm entry with the highest comprehensive score as the primary alarm entry, and other alarm entries as related alarm entries. The alarm generation time of the unified alarm task is the earliest time among all alarm entries, the risk level is the highest risk level, the geographical location correlation is the highest geographical location correlation, and the list of involved sensors is merged after removing duplicate device unique identifiers.

[0040] Unmerged alarm entries and unified alarm tasks are rearranged according to the initial processing order and sequentially written into the schedulable alarm task queue. Each alarm task records the task number, processing sequence number, main alarm entry, associated alarm entries, risk level, geographical location correlation, comprehensive score, alarm generation time, and a list of involved sensors in the queue. After the queue is written, subsequent dynamic scheduling and maintenance command generation directly read the alarm task queue, thereby avoiding duplicate task assignments for the same event by adjacent devices and prioritizing processing resources for alarm events with higher risk, stronger spatial correlation, and greater historical processing pressure.

[0041] Based on the alarm task queue, the scheduling model corresponding to the city's IoT platform is dynamically updated, and a corresponding set of operation and maintenance instructions is generated.

[0042] In this embodiment, the scheduling model is represented by a scheduling table, facility status parameters, and a sequence of instructions to be executed. The scheduling table records at least the task number, alarm location, risk level, geographical location correlation, task generation time, list of involved sensors, related facility numbers, preprocessing status identifier, and current occupancy status. The preprocessing status identifier includes four types: not preprocessed, pre-occupied, issued, and completed. The current occupancy status includes three types: idle, executing, and unavailable.

[0043] The first alarm task is read according to the processing order of the alarm task queue. The alarm location, risk level, geographical correlation, and list of involved sensors are extracted from this alarm task. The alarm location is used to determine related facilities. Related facilities are determined by calculating the distance between the alarm location and the facility's registered coordinates one by one. The distance is equal to the square root of the sum of the squares of the horizontal and vertical coordinate differences. When the calculated distance does not exceed the facility's registered service radius, the facility is determined as a related facility. If multiple facilities meet the criteria, the three closest facilities are retained for subsequent operation priority calculations; if fewer than three facilities meet the criteria, the actual number is retained.

[0044] The operational priority of each sensor and related facility is calculated based on the extracted information. Risk level conversion value = Risk level × 20. Geographical location correlation uses the 0-100 score obtained in the previous step. Task generation time sequence conversion value is determined according to the order of alarm tasks in the queue; the first alarm task receives 100 points, and subsequent alarm tasks are calculated as "100 - task sequence number × 5," with a score below 60 points. Facility distance conversion value is calculated as "1 - distance between facility and alarm location ÷ facility registered service radius" multiplied by 100, with a score below 0 points. Facility status conversion value is determined according to the current occupancy status: 100 points for idle, 50 points for running, and 0 points for unavailable. Operation score = Risk level conversion value × 35% + Geographical location correlation × 25% + Task generation time sequence conversion value × 15% + Facility distance conversion value × 15% + Facility status conversion value × 10%. The relevant facilities and related sensors are ranked from highest to lowest according to their operation scores to obtain the operation priority order. If the operation scores are the same, they are ranked from closest to furthest from the alarm location. After the ranking is completed, the preprocessing status of the facility corresponding to the alarm task in the scheduling table is updated from unpreprocessed to reserved, and the update time is recorded.

[0045] A set of operation and maintenance instructions is generated based on the operation priority and facility status identifier. When the relevant facility is unavailable, no equipment start or stop instructions are generated; only a field inspection request is generated, and the reason for unavailability is recorded. When the relevant facility is idle and the alarm task risk level is level 4 or 5, instructions to start the equipment, increase operating parameters, or activate backup facilities are generated. When the relevant facility is idle and the risk level is level 1 to 3, parameter adjustment and field inspection requests are generated. When the relevant facility is in execution, the remaining time of the executed task is read. If the remaining time is less than 10 minutes, a continuation execution instruction is generated; if the remaining time is more than 10 minutes, a field inspection request and backup facility call instruction are generated. Each operation and maintenance instruction records the task number, instruction type, execution object, execution order, target parameters, execution start time, and feedback requirements. The target parameters are determined based on the abnormal amplitude of the alarm entry. The parameter adjustment amount = equipment registered rated parameters × abnormal amplitude conversion score ÷ 100 × 20%. The adjusted target parameters must not exceed the equipment's registered safe operation limit.

[0046] The set of maintenance instructions is bound to alarm task entries. The binding includes the task number, the main alarm entry, associated alarm entries, and a list of involved sensors. After binding, the set of maintenance instructions is written into the sequence of instructions to be executed according to the alarm task queue order. Multiple maintenance instructions within the same alarm task are written sequentially according to their operation priority. After writing, the preprocessing status flag of the corresponding facility in the scheduling table is updated to "issued". The sequence of instructions to be executed is available for subsequent unified scheduling and dispatch, enabling high-risk alarms, spatially related alarms, and facility idle status to be considered synchronously within the same processing chain, reducing the situation of the same facility being called repeatedly or adjacent facilities being called out of order.

[0047] Based on the set of operation and maintenance instructions, all alarm events are uniformly scheduled and assigned, and the processing results and alarm data are fed back to the platform to iteratively update the alarm priority.

[0048] In this embodiment, the sequence of instructions to be executed is saved according to the task number, instruction type, execution object, execution order, target parameters, execution start time and feedback requirements, for unified scheduling and dispatch.

[0049] The system sequentially reads the maintenance instructions from the sequence of instructions to be executed. During reading, it first verifies that the task number matches the alarm task in the alarm task queue, and then executes the start, stop, parameter adjustment, or on-site inspection of the equipment involved in the corresponding alarm event according to the execution order. For instructions to start equipment, if the current occupancy status of the execution object is idle, the instruction is issued immediately; if the current occupancy status is in progress, the remaining execution time is compared with the follow-up waiting limit, with the follow-up waiting limit preferably set to 10 minutes. The rule is set to 25% of the shortest registered operation time for the facility; if the calculated result is less than 5 minutes, 5 minutes is used; if it exceeds 15 minutes, 15 minutes is used. If the remaining execution time does not exceed the follow-up waiting limit, the instruction is placed after the current operation; if the remaining execution time exceeds the follow-up waiting limit, the backup facility with the highest operation score in the same spatial area is called. For parameter adjustment instructions, the target parameters are executed according to the target parameters obtained in the previous step, and before execution, it is verified whether the registered safe operation limit of the equipment is exceeded; if it is exceeded, the instruction is issued according to the safe operation limit.

[0050] During execution, the execution time, operation completion status, and any abnormal situations for each instruction are recorded to form processing result data. Execution time includes the instruction issuance time, execution start time, and execution end time. Operation completion status includes four types: completed, incomplete, transferred to the site, and replaced. Abnormal situations include four categories: equipment unresponsive, inconsistent parameter feedback, communication interruption, and failure to reach the site for inspection. The optimal timeframe for determining equipment unresponsiveness is 3 minutes, set as 3 times the equipment's registered heartbeat cycle; if the calculated result is less than 1 minute, use 1 minute; if it exceeds 5 minutes, use 5 minutes. Inconsistent parameter feedback is calculated as "the absolute value of the difference between the target parameter and the feedback parameter ÷ the target parameter × 100%". When the result reaches 5%, it is considered inconsistent parameter feedback; 5% is the preferred value. For precision metrology equipment, 2% can be used; for start / stop control equipment, 8% can be used.

[0051] The processed data and the corresponding original alarm event data are merged according to the task number, and the processing history of the alarm event is updated. The response time equals the start time of execution minus the alarm generation time; the processing time equals the end time of execution minus the start time of execution. The completion status is determined according to the processed data: 1 for completed, 0 for incomplete, 0.5 for transferred to the site, and 0.8 for replaced. The impact range correction value is calculated based on the change in the number of sensors involved in the alarm entry before and after processing: Impact range correction value = number of sensors still abnormal after processing ÷ number of abnormal sensors before processing × 100; when the number of abnormal sensors before processing is 0, the impact range correction value is 0. The closer the impact range correction value is to 0, the less residual impact after processing. After merging, the processing history should include at least the task number, response time, processing time, completion status, abnormal situation, impact range correction value, and update time.

[0052] The comprehensive score is recalculated based on the updated processing history, risk level, and geographic location correlation. The risk level conversion value remains the same: risk level multiplied by 20. Feedback score = response time score × 40% + completion status score × 35% + impact range score × 25%. Response time score = response time ÷ prescribed response time × 100, rounded down to 100 if exceeding 100; completion status score = [1 - completion status] × 100; impact range score = impact range correction value. Updated comprehensive score = risk level conversion value × 45% + geographic location correlation × 25% + feedback score × 30%. When the same alarm event is completed by a replacement facility, the backup facility number is written into the processing history and participates in the calculation with the completion status corresponding to the replaced status. After recalculation, incomplete and transferred alarm tasks are reordered from high to low according to the updated comprehensive score; completed alarm tasks are removed from the alarm task queue. Through this process, high-risk alarm events, those with timeouts, and those with residual impacts after handling can be prioritized in the next round of dispatching.

[0053] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application.

Claims

1. An alarm management method based on a city IoT platform, characterized in that, include: Access multiple infrastructure sensors and devices in the city's Internet of Things (IoT) platform to obtain real-time operational data from each device; Anomaly detection is performed on the acquired real-time data to generate an initial set of alarm information, and the risk level and geographical location correlation of each alarm are calculated. The initial alarm information set is used to construct an alarm priority model based on risk level, correlation, and historical processing records, forming a schedulable alarm task queue. Based on the alarm task queue, the scheduling model corresponding to the city's IoT platform is dynamically updated, and a corresponding set of operation and maintenance instructions is generated. Based on the set of operation and maintenance instructions, all alarm events are uniformly scheduled and assigned, and the processing results and alarm data are fed back to the platform to iteratively update the alarm priority.

2. The alarm management method based on a city IoT platform according to claim 1, characterized in that, Anomaly detection is performed on the acquired real-time operational data, including the following steps: The real-time operating data of the same sensor within the same time slice are sorted according to the acquisition time, and data marked as missing or with communication delay are removed to obtain a valid numerical sequence. The short-term amplitude range is calculated based on the maximum and minimum monitoring values ​​within the effective numerical sequence, and the continuous stable duration is accumulated based on the changes in adjacent effective values. Based on the combination relationship between short-term amplitude range and continuous stable duration, abnormal information is classified into sudden abnormalities or slow drift abnormalities, thus obtaining abnormal information with abnormality type.

3. The alarm management method based on a city IoT platform according to claim 2, characterized in that, The initial alarm information set is generated by the following steps: calculating the temporal overlap and spatial proximity of abnormal information in the same spatial area, and obtaining the spatiotemporal overlap value by multiplying the temporal overlap and spatial proximity; taking the earliest generated abnormal information as the starting point, classifying abnormal information with a spatiotemporal overlap value of 0.6 into the same abnormal event; and forming preliminary alarm entries based on the abnormal type, center location, earliest acquisition time, and list of involved sensors in the same abnormal event.

4. The alarm management method based on a city IoT platform according to claim 1, characterized in that, Calculating the risk level of each alarm involves the following steps: The abnormal amplitude is calculated based on the absolute value of the difference between the actual monitored value and the normal reference value and the allowable deviation value. The normal reference value is the median of the effective monitored values ​​of the same equipment within the same time period. Based on the number of sensors involved in the abnormal event, the duration of the abnormal event, and the frequency of similar historical events, the abnormal amplitude conversion score, the spatiotemporal concentration conversion score, and the historical event frequency conversion score are obtained respectively, and the risk level is determined according to the risk score.

5. The alarm management method based on a city IoT platform according to claim 3, characterized in that, Calculate the geographic location correlation for each alarm, including the following steps: The alarm entries in the initial alarm information set are compared according to geographical coordinates and timestamps, and the alarm entries whose spatial distance does not exceed the registered service radius of the relevant facilities and whose alarm time difference does not exceed the specified association duration are selected as the neighboring alarm points; The density and service overlap ratio are obtained based on the number of nearby alarm points, the service area of ​​related facilities, and the overlapping service area. Based on the distance from the alarm entry to the center of the event cluster, the density, and the service overlap ratio, calculate the geographical location correlation and attach it to the corresponding alarm entry.

6. The alarm management method based on a city IoT platform according to claim 1, characterized in that, To create a schedulable alarm task queue, the following steps are included: The risk level is converted into a risk level conversion value, and a comprehensive score is calculated by combining the geographical location correlation and historical processing record scores. The historical processing record scores are obtained based on the average response time and number of times the same anomaly type is processed in the same spatial area. All alarm items are sorted from high to low according to the comprehensive score, and if the comprehensive scores are the same, they are sorted from early to late according to the alarm generation time to obtain the processing order of alarm tasks.

7. The alarm management method based on a city IoT platform according to claim 6, characterized in that, Creating a schedulable alarm task queue also includes the following steps: For alarm entries in the same spatial area whose comprehensive score reaches the merging reference value, calculate the spatial proximity value and the temporal proximity value; calculate the merging judgment value according to the spatial proximity value accounting for 60% and the temporal proximity value accounting for 40%, and merge alarm entries with a merging judgment value of 0.7 into a unified alarm task; take the alarm entry with the highest comprehensive score as the main alarm entry, take the remaining alarm entries as associated alarm entries, and write the unified alarm task into the alarm task queue.

8. The alarm management method based on a city IoT platform according to claim 1, characterized in that, Generate the corresponding set of operation and maintenance instructions, including the following steps: Read the first alarm task in the alarm task queue, and determine no more than 3 related facilities based on the distance between the alarm location and the facility's registered coordinates; calculate the operation score based on the risk level conversion value, geographical location correlation, task generation time sequence conversion value, facility distance conversion value, and facility status conversion value; determine the operation priority order of related facilities and sensors involved according to the operation score from high to low, and update the preprocessing status mark of the corresponding facility in the scheduling table to "pre-occupied".

9. The alarm management method based on a city IoT platform according to claim 1, characterized in that, Generating the corresponding set of operation and maintenance instructions also includes the following steps: Based on the current occupancy status and risk level of relevant facilities, generate operation and maintenance instructions to start equipment, stop equipment, adjust parameters, call backup facilities, or send on-site inspection requests; the parameter adjustment amount is calculated according to the equipment's registered rated parameters, abnormal amplitude conversion score, and 20% adjustment ratio, and the adjusted target parameters do not exceed the equipment's registered safe operation limit; bind the set of operation and maintenance instructions to alarm task entries, and write them into the sequence of instructions to be executed according to the order of the alarm task queue.

10. The alarm management method based on a city IoT platform according to claim 9, characterized in that, Iterative updates to alarm priorities include the following steps: The system sequentially reads the maintenance instructions from the sequence of instructions to be executed, records the instruction issuance time, execution start time, execution end time, operation completion status, and any abnormal situations, and forms processing result data. The processing result data is then merged with the original data of the corresponding alarm events according to the task number to obtain a processing history record containing response time, processing time, completion status, and impact range correction values. Based on the updated processing history record, risk level, and geographical location correlation, the comprehensive score is recalculated, and incomplete and transferred alarm tasks are reordered, while completed alarm tasks are removed from the alarm task queue.

Images