Urban lifeline engineering iot perception device detection and data aggregation platform

Abstract

The present application relates to the field of smart city, especially to a kind of city lifeline engineering IOT sensing device detection and data aggregation platform, comprising: protocol analysis module, it is used to convert original sensing data into the standard data of uniform format;Monitoring analysis module, it is used to determine whether standard data is disturbance data according to the data integrity and timing fluctuation degree of the standard data;Regional division module, it is used to determine whether to adjust the regional selection reference according to the source distribution stability of disturbance data, to divide the device focusing area based on regional selection reference;Dynamic processing module, it is used to construct several regional allocation sets according to spatial distance, and determine whether to perform task offloading according to the model requirement update rate of processing unit;The dynamic processing module includes several processing units, which are used to analyze and process using pre-trained model.The present application improves the real-time, accuracy and system operation stability of city lifeline operation early warning.

Description

Technical Field

[0001] This invention relates to the field of smart cities, and in particular to a platform for detecting and aggregating data from IoT sensing devices in urban lifeline engineering projects. Background Technology

[0002] Urban lifeline engineering refers to critical infrastructure systems that maintain the normal operation of cities and meet the production and living needs of the public. It encompasses multiple key areas such as gas, bridges, tunnels, water supply, drainage, heating, electricity, and integrated utility tunnels. Statistics show that the total length of underground pipelines nationwide now exceeds 3 million kilometers, and the total number of bridges exceeds 1 million, leading to an ever-increasing demand for urban lifeline engineering. The need for safety assurance in urban lifeline engineering is becoming increasingly urgent. Currently, urban lifeline safety engineering is utilizing modern information technologies such as the Internet of Things, big data, cloud computing, and artificial intelligence to conduct real-time monitoring, dynamic early warning, precise tracing, and collaborative handling of the operational status of urban infrastructure, in order to identify and control potential risks as early as possible and effectively prevent accidents. The vast majority of provinces and regions across the country have established unified provincial-level monitoring platforms, and all prefecture-level and above cities have initiated the implementation of lifeline engineering projects. However, in existing technologies, urban lifeline engineering projects suffer from large-scale monitoring equipment deployments and relatively fixed monitoring areas during the initial installation phase, making it difficult to respond promptly to dynamic changes in the operating environment. Data from monitoring devices occasionally malfunctions, and the concentrated influx of localized sensing data leads to a surge in data center processing pressure. Simultaneously, the optimization methods for model calculations lack flexibility, making it difficult to adapt to dynamic fluctuations in data distribution and computational load. These combined deficiencies severely impact the real-time performance of lifeline risk warnings and the stability of system operation, resulting in predictions often accompanied by a high false alarm rate. Therefore, improving the accuracy and reliability of urban lifeline warnings is a significant concern for those skilled in the art.

[0003] Chinese Patent Publication No. CN122022442A discloses a method, apparatus, computer equipment, and storage medium for comprehensive risk analysis and assessment of urban lifelines. The method includes: establishing urban lifeline safety engineering data standards; determining comprehensive risk indicators based on arbitrary regions according to the urban lifeline safety engineering data standards; classifying the comprehensive risk indicators according to three elements: risk source, disaster-bearing body, and disaster mitigation capacity; further dividing all determined comprehensive risk indicators according to their status and assigning risk values ​​based on the urban lifeline safety engineering data standards; calculating a risk index, assessment index, and comprehensive risk level based on the risk values; and judging the risk status of any region based on the comprehensive risk level. However, the above technical solution suffers from the following problems: it cannot determine the data type based on actual data changes in the data acquisition terminal to obtain a model processing method for comprehensive data aggregation, which easily leads to poor accuracy in the actual early warning results for urban lifelines. Summary of the Invention

[0004] This invention provides a detection and data aggregation platform for IoT sensing devices in urban lifeline engineering projects. It overcomes the problem in existing technologies where the model processing method cannot determine the data type based on the actual data changes in the collection terminal to obtain comprehensive data aggregation information, which easily leads to poor accuracy of actual early warning results for urban lifelines.

[0005] Therefore, the present invention provides a platform for the detection and data aggregation of IoT sensing devices for urban lifeline engineering projects, comprising: The protocol parsing module is used to access various IoT sensing devices in the urban lifeline project. Based on the communication protocol type of each device, it calls the corresponding protocol parser to convert the raw sensing data into standard data in a unified format. The monitoring and analysis module is used to determine whether the standard data is disturbed data based on the data integrity and time series fluctuation of the standard data. The region division module is used to determine whether to adjust the region selection benchmark from urban distribution characteristics to interference distribution characteristics based on the stability of the source distribution of disturbance data, so as to divide the equipment focusing area based on the region selection benchmark; The dynamic processing module is used to construct several regional allocation sets based on spatial distance, and to determine whether to unload tasks based on the model requirement update rate of the processing unit; the dynamic processing module includes several processing units, which are used to analyze and process standard data using pre-trained models.

[0006] Furthermore, the monitoring and analysis module includes a disturbance analysis unit, which is used to identify standard data with a data integrity less than a preset data integrity or a time series fluctuation greater than a preset time series fluctuation, and determine it as disturbance data.

[0007] Furthermore, the monitoring and analysis module also includes a data analysis unit, which is used to determine the data integrity based on the quantity of key data types and to determine the time series fluctuation based on the upload time difference of adjacent standard data combinations. The data completeness is positively correlated with the number of key data types. The time-series volatility is positively correlated with the upload time difference of adjacent standard data combinations.

[0008] Furthermore, the region division module includes a statistical analysis unit and a division selection unit; The statistical analysis unit is used to determine the source distribution stability based on the spatiotemporal change rate of the IoT sensing devices corresponding to the disturbance data; The selection unit is divided to respond to the condition that the source distribution stability is greater than the preset source distribution stability, and to determine whether to adjust the regional selection benchmark from urban distribution characteristics to interference distribution characteristics.

[0009] Furthermore, the region division module also includes a division execution unit, which is used to respond to the condition that the region selection benchmark is the interference distribution characteristics, and to construct regions for the IoT sensing devices with disturbance data in descending order of stable sub-values, so as to obtain a number of device focus areas that meet the preset construction conditions. The preset construction condition is that the average value of the stable sub-values ​​of IoT sensing devices within the device's focal area is less than or equal to the preset average stable sub-value.

[0010] Furthermore, the partitioning execution unit is also used to respond to the condition that the regional selection criterion is urban distribution characteristics, and to construct regions for the IoT sensing devices with disturbance data in descending order of regional building density, so as to obtain the device focus area.

[0011] Furthermore, the dynamic processing module includes a space allocation unit, which is used to allocate the focus area of ​​each device in ascending order of spatial distance to obtain several set of area allocations; The regional allocation set includes processing units and several device focus areas.

[0012] Furthermore, the dynamic processing module also includes an adaptive analysis unit, which is used to determine the task offloading for the region allocation set whose model demand update rate is greater than the preset model demand update rate based on the condition of the model demand update rate of the processing unit. The task offloading target is the IoT sensing device with the largest disturbance frequency.

[0013] Furthermore, the dynamic processing module also includes a benchmark analysis unit, which is used to determine to adjust the task unloading benchmark from task uniformity to running interval compliance in response to the condition that the task diversity balance of the processing unit is greater than the preset task diversity balance.

[0014] Furthermore, the dynamic processing module also includes an unloading allocation unit, which is used to migrate the unloaded IoT sensing devices to the area allocation set corresponding to the processing unit with the largest task unloading benchmark.

[0015] Compared with the prior art, the beneficial effects of the present invention are that the present invention determines whether the data is disturbed based on the data integrity and time series fluctuation of the standard data through the monitoring and analysis module. The data integrity reflects the completeness of the key information fields contained in the standard data uploaded in a single upload, and the time series fluctuation reflects the degree to which the time interval of the standard data deviates from the normal rhythm in the upload sequence of its respective device. When data is missing or the upload rhythm is abnormal, disturbed data can be identified in a timely manner, avoiding misjudgment and omission due to incomplete data or abnormal transmission, and improving the accuracy and real-time performance of the perceived data quality monitoring.

[0016] Furthermore, this invention adaptively determines the region selection benchmark based on the source distribution stability of the disturbance data through a region division module. The source distribution stability reflects the overall level of the spatiotemporal change rate of the IoT sensing devices corresponding to each disturbance data. The greater the spatiotemporal change rate, the more unstable the temporal sequence of the device disturbance. When the source distribution stability is high, the region selection benchmark is adjusted from urban distribution characteristics to disturbance distribution characteristics. The device focusing area is constructed based on stable sub-values, avoiding the problem of inaccurate focusing area division caused by using a fixed benchmark when the urban functional zoning and disturbance source distribution are inconsistent. This improves the pertinence and rationality of the device focusing area division.

[0017] Furthermore, this invention constructs a region for the IoT sensing device that is disturbing data by sequentially constructing the region in descending order of building density to obtain the device focus area. The regional building density reflects the area covered by buildings within the initial construction range of the device. This can prioritize the inclusion of devices with dense surrounding buildings and high sensing needs into the focus area, ensuring the integrity of device coverage in high-risk areas of urban lifeline projects and avoiding blind spots in safety monitoring caused by high building density but devices not being included in the focus area.

[0018] Furthermore, this invention uses a dynamic processing module to determine whether to unload tasks based on the model demand update rate of the processing unit. The model demand update rate reflects the frequency of model demand changes in a single processing unit. A model demand change is triggered by a continuous increase in the monitored value within the device's focus area. When the model demand update rate is high, the device with the highest disturbance frequency is unloaded and migrated to other processing units. This avoids the problem of decreased processing efficiency due to excessive load on a single processing unit or excessively frequent model updates, and achieves dynamic and balanced allocation of computing resources, thereby improving the overall data processing timeliness of the platform.

[0019] Furthermore, this invention determines the task offloading benchmark based on the task diversity balance of the processing unit through the benchmark analysis unit. The task diversity balance reflects the degree of balance in the number of calling models in each processing unit. The offloaded devices are migrated to the processing unit with the largest offloading benchmark, avoiding the problem that a single offloading benchmark cannot effectively distinguish the priority of device migration when the task distribution is balanced. This achieves the optimal matching between the offloading target and the current load characteristics of the processing unit, further improving the rationality of task offloading and resource utilization efficiency. Attached Figure Description

[0020] Figure 1 This is a module structure diagram of the IoT sensing device detection and data aggregation platform for urban lifeline engineering according to an embodiment of the present invention; Figure 2 A flowchart for determining disturbance data in an embodiment of the present invention; Figure 3This is a flowchart illustrating whether to adjust the region selection criterion in an embodiment of the present invention; Figure 4 This is a flowchart for determining whether to perform task unloading in an embodiment of the present invention. Detailed Implementation

[0021] To make the objectives and advantages of the present invention clearer, the present invention will be further described below with reference to embodiments; it should be understood that the specific embodiments described herein are merely for explaining the present invention and are not intended to limit the present invention.

[0022] Preferred embodiments of the present invention will now be described with reference to the accompanying drawings. Those skilled in the art should understand that these embodiments are merely illustrative of the technical principles of the present invention and are not intended to limit the scope of protection of the present invention.

[0023] It should be noted that in the description of this invention, the terms "upper", "lower", "left", "right", "inner", "outer", etc., which indicate directions or positional relationships, are based on the directions or positional relationships shown in the accompanying drawings. This is only for the convenience of description and is not intended to indicate or imply that the device or element must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, it should not be construed as a limitation of this invention.

[0024] Please see Figure 1 The diagram shown is a modular structure diagram of the IoT sensing device detection and data aggregation platform for urban lifeline engineering projects according to an embodiment of the present invention. This embodiment of the present invention provides an IoT sensing device detection and data aggregation platform for urban lifeline engineering projects, comprising: The protocol parsing module is used to access various IoT sensing devices in the urban lifeline project. Based on the communication protocol type of each device, it calls the corresponding protocol parser to convert the raw sensing data into standard data in a unified format. The monitoring and analysis module, which is connected to the protocol parsing module, is used to determine whether the standard data is disturbed data based on the data integrity and time series fluctuation of the standard data. The area division module, which is connected to the monitoring and analysis module, is used to determine whether to adjust the area selection benchmark from urban distribution characteristics to interference distribution characteristics based on the stability of the source distribution of disturbance data, so as to divide the equipment focusing area based on the area selection benchmark. The dynamic processing module, which is connected to the region partitioning module and the protocol parsing module respectively, is used to construct several region allocation sets based on spatial distance and determine whether to unload tasks based on the model requirement update rate of the processing unit. The dynamic processing module includes several processing units, which are used to analyze and process standard data using pre-trained models.

[0025] In this embodiment of the invention, the process by which the protocol parsing module obtains the locations of each building and IoT sensing device includes: the protocol parsing module obtaining a digital twin model of the monitoring area to obtain the locations of each building and IoT sensing device. The monitoring area is the smallest rectangle that can contain all the buildings that need to be monitored in the urban lifeline project. A three-dimensional spatial model is constructed using urban geographic information system data, building information model data, and IoT sensing device deployment maps. This model integrates the spatial location data, geometric shape data, and device deployment data of urban infrastructure to form a digital spatial mapping that includes the outline, height, type of each building, and the coordinates of each IoT sensing device. The method for constructing the digital twin model is existing technology in this field and will not be described in detail here.

[0026] In this embodiment of the invention, the process by which the protocol parsing module acquires standard data from various IoT sensing devices includes: the protocol parsing module monitors the operational status and raw sensing data of the infrastructure by connecting to IoT sensing devices in various infrastructures along the city's lifeline; it connects to the communication protocol interfaces of each IoT sensing device, reads the communication protocol type of each IoT sensing device, and calls the corresponding protocol parser. For example, for low-power IoT sensing devices such as gas valves and underground pipe networks, combustible gas monitors and pressure sensors can be used, employing NB-IoT narrowband IoT communication for transmission; for IoT sensing devices requiring real-time monitoring, such as water pipe network monitoring, level gauges and water quality monitors can be used, employing 4G Cat.1 communication for transmission; and for IoT sensing devices requiring stable connectivity, such as bridges, inclinometers and displacement sensors can be used, employing Ethernet for transmission. The moment when each piece of raw sensing data is uploaded to the protocol parsing module via the IoT sensing device is recorded as the upload time. The protocol parsing module then converts the raw sensing data according to a unified data format to generate standard data. The standard data from each IoT sensing device within the detection period are arranged sequentially from earliest to latest upload time to obtain the upload sequence for each IoT sensing device. Each piece of standard data in the upload sequence carries a timestamp of its upload time. The protocol parsing module, based on the device type and communication protocol used by each IoT sensing device, calls the corresponding protocol parser to unpack, extract fields, and restructure the format of the raw sensing data, converting it into unified structured data containing device identification, monitoring values, data types, collection timestamps, and device location information.

[0027] Key data within the standard data includes, but is not limited to, the name, data type, monitoring values, data collection timestamp, and device location of the IoT sensing devices. The data types include, but are not limited to, the pressure of gas pipelines, the concentration of flammable gases in adjacent underground spaces, the flow rate of water supply networks, the crack area in images of bridge structural crack detection, and the area of ​​accumulated water in videos of urban road flooding monitoring.

[0028] In this embodiment of the invention, the protocol parsing module acquires standard data from each IoT sensing device within a cyclical detection period. The acquisition process includes: simulating the operation of multiple sets of historical operating data from IoT sensing devices with different detection period lengths; setting different detection period length thresholds for disturbance data identification and device focus area division tests; and selecting the average value of the detection period length thresholds that meet user requirements for the average response delay of disturbance data identification as the preset detection period length. It can be understood that the average response delay for disturbance data identification is the average time interval between the IoT sensing device uploading raw sensing data and the platform determining whether the data is disturbance data. The shorter the average response delay, the stronger the platform's real-time monitoring capability, and the greater the user's sensitivity to the platform's real-time monitoring capability. Therefore, when selecting a preset value using the average response delay for disturbance data identification as an evaluation index, the requirement for the average response delay is lower; that is, the shorter the average response delay, the better it meets user needs. The setting of the evaluation index is content already understood by those skilled in the art and will not be elaborated further. In this embodiment, the longer the detection period, the more standard data is acquired, and the higher the accuracy of perturbation data judgment. The more sensitive the user is to the accuracy of the device's detection data acquisition, the longer the detection period should be.

[0029] Please see Figure 2 As shown, it is a flowchart for determining disturbance data in an embodiment of the present invention. The monitoring and analysis module includes a disturbance analysis unit, which is used to identify standard data with a data integrity less than a preset data integrity or a time series fluctuation greater than a preset time series fluctuation, and determine it as disturbance data.

[0030] Specifically, the monitoring and analysis module also includes a data analysis unit, which is used to determine the data integrity based on the quantity of key data types and to determine the time series fluctuation based on the upload time difference of adjacent standard data combinations. The data completeness is positively correlated with the number of key data types. The time-series volatility is positively correlated with the upload time difference of adjacent standard data combinations.

[0031] In this embodiment of the invention, the monitoring and analysis module further includes a data calculation unit for calculating data integrity and time-series fluctuation. The process of calculating data integrity includes: determining data integrity based on the number of key data types contained in the standard data within the most recent complete detection period, wherein the data integrity is positively correlated with the minimum number of key data types. In this embodiment, the data integrity is calculated as the minimum number of key data types contained in the standard data within the most recent complete detection period. It can be understood that data integrity is used to determine the completeness of key information fields contained in a single uploaded standard data set. The higher the data integrity, the more key data types the standard data contains, and the more fully the information is carried.

[0032] The process of obtaining temporal volatility includes: the temporal volatility is determined based on the upload time difference of adjacent standard data combinations corresponding to the standard data and the average upload time difference of each adjacent standard data combination in the corresponding upload sequence. The temporal volatility is positively correlated with the upload time difference and negatively correlated with the average value. In this embodiment, the temporal volatility is the ratio of the upload time difference of the standard data and the adjacent standard data combination formed by the standard data and the nearest previous standard data in the upload sequence of the standard data within the most recent complete detection period to the average upload time difference of each adjacent standard data combination in the corresponding upload sequence of the standard data. The adjacent standard data combination consists of the upload times of two adjacent standard data in the upload sequence of the IoT sensing device corresponding to the standard data. The upload time difference is the absolute value of the difference between the two upload times in the adjacent standard data combination. It is understood that such irregularly uploaded standard data is difficult to form a continuous or equally spaced upload sequence with the standard data collected before and after by the same IoT sensing device, resulting in the pre-trained model being unable to establish accurate temporal correlation features when using this upload sequence for temporal prediction. At this point, the monitoring and analysis module marks the standard data with low data integrity or high time series fluctuation, so that the region division module can dynamically adjust the division benchmark of the device's focus area according to the source distribution of the disturbance data. This allows the subsequent processing unit to focus only on the stable standard data for model analysis to obtain more accurate and faster prediction results.

[0033] In this embodiment of the invention, obtaining the preset time-series volatility includes: simulating multiple sets of standard data from IoT sensing devices with different time-series volatility; setting different time-series volatility thresholds for perturbation data judgment tests; and selecting the average value of the time-series volatility thresholds whose recall rate for perturbation data judgment meets user requirements as the preset time-series volatility. The recall rate for perturbation data judgment is the proportion of standard data correctly judged as perturbation data to the total number of actual perturbation data. The greater the user's sensitivity to the risk of missing perturbation data, the higher the requirement for recall rate when selecting the preset value using the recall rate for perturbation data judgment as the evaluation index. The higher the user's sensitivity to the risk of missing perturbation data, the smaller the value of the preset time-series volatility.

[0034] In this embodiment of the invention, obtaining the preset data integrity includes: simulating the data by collecting multiple sets of standard data from IoT sensing devices with different data integrity levels; setting different data integrity thresholds for perturbation data judgment tests; and selecting the average value of the data integrity thresholds whose accuracy in perturbation data judgment meets user requirements as the preset data integrity. In this embodiment, the more sensitive the user is to the accuracy of perturbation data identification, the larger the preset data integrity value.

[0035] Please see Figure 3 As shown, it is a flowchart of whether to adjust the region selection benchmark in an embodiment of the present invention. The region division module includes a statistical analysis unit and a division selection unit. The statistical analysis unit is used to determine the source distribution stability based on the spatiotemporal change rate of the IoT sensing devices corresponding to the disturbance data; The selection unit is divided to respond to the condition that the source distribution stability is greater than the preset source distribution stability, and to determine whether to adjust the regional selection benchmark from urban distribution characteristics to interference distribution characteristics.

[0036] The region division module also includes a stability calculation unit for calculating the spatiotemporal change rate and source distribution stability of IoT sensing devices. The process of obtaining the spatiotemporal change rate includes: the spatiotemporal change rate is determined based on the ratio of the variance to the average of the upload time interval of the disturbance data combination of the IoT sensing device in each detection period within the most recent complete historical acquisition period. The spatiotemporal change rate is positively correlated with the variance and negatively correlated with the average. In this embodiment, for the IoT sensing device corresponding to a single disturbance data, the spatiotemporal change rate is the ratio of the variance to the average of the upload time interval of the disturbance data combination of the IoT sensing device in each detection period within the most recent complete historical acquisition period. The disturbance data combination consists of two temporally adjacent disturbance data within the most recent complete historical acquisition period, and the upload time interval is the time interval between the upload times of the two disturbance data in the disturbance data combination. Understandably, the spatiotemporal change rate is used to determine the degree of dispersion of the time interval of disturbance data occurrence for a single IoT sensing device within the historical collection period. The greater the spatiotemporal change rate, the more unpredictable the disturbance characteristics of the IoT sensing device are. Static urban distribution characteristics can no longer effectively reflect the actual disturbance distribution. At this time, the disturbance distribution characteristics should be switched to make the division of the device's focus area match the changes of the IoT sensing device corresponding to the actual disturbance data.

[0037] The process of obtaining source distribution stability includes: the source distribution stability is determined based on the average spatiotemporal change rate of each IoT sensing device within a historical acquisition period, and the source distribution stability is positively correlated with the average spatiotemporal change rate. In this embodiment, the source distribution stability is the average spatiotemporal change rate of each IoT sensing device within a historical acquisition period. It can be understood that source distribution stability is used to determine the degree of distribution consistency of IoT sensing devices corresponding to perturbation data. The higher the source distribution stability, the more similar the dispersion of IoT sensing devices corresponding to each perturbation data, indicating that the distribution in the time dimension is highly discrete and lacks a clear spatiotemporal clustering pattern.

[0038] In this embodiment of the invention, obtaining the preset source distribution stability includes: simulating disturbance data from multiple IoT sensing devices with different source distribution stability values; setting different source distribution stability thresholds for regional selection benchmark adjustment and judgment tests; and selecting the average value of the source distribution stability thresholds whose average absolute error in disturbance source distribution identification meets user requirements as the preset source distribution stability. In this embodiment, the average absolute error in disturbance source distribution identification is the average of the absolute values ​​of the differences between the calculated value and the true value of the disturbance source distribution stability. The greater the user's sensitivity to the accuracy of disturbance source distribution identification, the lower the requirement for the average absolute error and the smaller the value when selecting the preset value using the average absolute error of disturbance source distribution identification as an evaluation index. The more sensitive the user is to the identification of disturbance source aggregation areas, the smaller the preset source distribution stability value.

[0039] In this embodiment of the invention, a historical data collection period for standard data over a historical period is defined. The acquisition process includes: simulating the operation of multiple sets of historical data from IoT sensing devices with different historical collection period lengths; setting different historical collection period thresholds for disturbance source distribution identification tests; and selecting the average value of the historical collection period thresholds whose average absolute error in disturbance source distribution identification meets user requirements as the historical collection period. In this embodiment, the more sensitive the user is to the accuracy of disturbance source distribution identification, the larger the historical collection period value.

[0040] Specifically, the region division module also includes a division execution unit, which is used to respond to the condition that the region selection benchmark is the interference distribution characteristics, and to construct regions for the IoT sensing devices with disturbance data in descending order of stable sub-values, so as to obtain a number of device focus areas that meet the preset construction conditions. The preset construction condition is that the average value of the stable sub-values ​​of IoT sensing devices within the device's focal area is less than or equal to the preset average stable sub-value.

[0041] The region division module also includes a sub-value calculation unit, which is used to obtain the stable sub-values ​​of IoT sensing devices that generate perturbation data and the average of the stable sub-values ​​of two IoT sensing devices. The acquisition process includes: the stable sub-value is determined based on the number of perturbation data and the number of standard data in the IoT sensing device within the most recent complete historical acquisition cycle, and the stable sub-value and the ratio of the number of perturbation data are negatively correlated. In this embodiment, the stable sub-value is the difference between the ratio of the number of perturbation data to the number of standard data in the IoT sensing device within the most recent complete historical acquisition cycle. It is worth noting that if the number of standard data in the most recent complete historical acquisition cycle is zero, the stable sub-value is 0, indicating that the IoT sensing device has no valid data uploaded within the historical acquisition cycle and does not have the reliability for region construction. It can be understood that the stable sub-value is used to determine the stability of the IoT sensing device's output of normal standard data within the historical acquisition cycle. The larger the stable sub-value, the lower the proportion of perturbation data to standard data in the IoT sensing device, the higher the stability of the device's output data, and the more suitable it is as the core for constructing the device's focused region.

[0042] The process of dividing the execution unit to obtain the device focus area based on the interference distribution characteristics includes: when constructing a region for an IoT sensing device with a single disturbance data, a preset number of IoT sensing devices closest to that IoT sensing device are identified and recorded as initial devices. The preset number is the ratio of the number of IoT sensing devices in the monitoring area to the number of IoT sensing devices in the disturbance data. If the average of the stable sub-values ​​of the IoT sensing device and the initial devices is less than or equal to a preset average stable sub-value, then the smallest rectangular area that can surround the IoT sensing device and the initial devices is recorded as a device focus area; otherwise, the IoT sensing devices with the disturbance data are sequentially eliminated according to the order of distance from the initial devices to the IoT sensing devices with the disturbance data from farthest to closest, and the average of the stable sub-values ​​is calculated until the average of the stable sub-values ​​of the IoT sensing device and the initial devices is less than or equal to the preset average stable sub-value. Then, the smallest rectangular area that can surround the IoT sensing devices with the disturbance data and each initial device is recorded as a device focus area. It is worth noting that if an IoT sensing device with disturbed data has already been assigned to a certain device focus area, then no further region construction will be carried out for that IoT sensing device; if, after the device focus area corresponding to the IoT sensing device with disturbed data has been constructed, there are IoT sensing devices that have not participated in the construction of the device focus area, then they will be assigned to the nearest device focus area.

[0043] In this embodiment of the invention, obtaining the preset average stability sub-value includes: simulating the distribution data of IoT sensing devices with multiple sets of different average stability values; setting different average stability thresholds for device focus area construction tests; and selecting the average value of the average stability thresholds that satisfy user requirements for the recall rate of the constructed device focus area as the preset average stability sub-value. In this embodiment, the recall rate of the constructed device focus area is the proportion of the number of stable devices correctly included in the device focus area to the total number of stable devices that should be included in the device focus area. The greater the user's sensitivity to the completeness of the coverage of the constructed device focus area, the higher the requirement for the recall rate when selecting the preset value using the recall rate of the constructed device focus area as the evaluation index. The higher the user's requirement for the completeness of the coverage of stable devices within the device focus area, the smaller the value of the preset average stability sub-value.

[0044] Specifically, the partitioning execution unit is also used to respond to the condition that the regional selection benchmark is urban distribution characteristics, and to construct regions for IoT sensing devices with disturbance data in descending order of regional building density, so as to obtain the device focus area.

[0045] The region segmentation module also includes a density calculation unit, which is used to acquire the regional building density of IoT sensing devices that generate disturbance data. The acquisition process includes: the regional building density is determined based on the area covered by buildings within the initial construction range of a single IoT sensing device and the area of ​​the initial construction range. The regional building density is positively correlated with the building coverage area and negatively correlated with the area of ​​the initial construction range. In this embodiment, for an IoT sensing device that generates disturbance data, the regional building density is the ratio of the area covered by buildings within the initial construction range of the IoT sensing device to the area of ​​the initial construction range. The initial construction range is a circular area with the location of the IoT sensing device as the center and the construction distance as the radius. The buildings include, but are not limited to, residential buildings, bridges, and tunnels. It can be understood that the regional building density is used to determine the spatial density of the buildings surrounding the location of the IoT sensing device. The higher the regional building density, the higher the proportion of the surrounding building coverage area, the more significant the urban structural characteristics, and the denser the sensing demand.

[0046] The process of dividing the execution unit to obtain the device focus area based on urban distribution characteristics includes: for an IoT sensing device with a single disturbance data point, each IoT sensing device within the construction distance of that IoT sensing device is recorded as an analysis device. This process is repeated until no new analysis devices appear within the construction distance, at which point the loop stops. The smallest rectangular area that can surround the IoT sensing device with the disturbance data and its corresponding analysis device is recorded as a device focus area. This area construction continues until all IoT sensing devices within the monitoring area are included in the device focus area.

[0047] The process of obtaining the build distance by dividing the execution unit includes: simulating the spatial distribution data of IoT sensing devices at multiple sets of different build distances; setting different build distance thresholds for device focus area division tests; and selecting the average value of the build distance thresholds whose recall rate for device focus area division meets user requirements as the preset build distance. In this embodiment, the recall rate of the device focus area division is the proportion of the number of IoT sensing devices correctly assigned to the device focus area to the total number of IoT sensing devices that should be assigned to the device focus area. The greater the user's sensitivity to the completeness of device focus area coverage, the higher the requirement for recall rate when selecting the preset value using the recall rate of device focus area division as the evaluation index. The higher the user's requirement for the completeness of device focus area coverage, the larger the value of the preset build distance.

[0048] Specifically, the dynamic processing module includes a space allocation unit, which is used to allocate the focus area of ​​each device based on the ascending spatial distance to obtain a set of area allocations. The regional allocation set includes processing units and several device focus areas.

[0049] The process of acquiring a region allocation set by the spatial allocation unit includes: for a single processing unit, the spatial distance is the distance between its location coordinates and the coordinates of the center point of the area of ​​each device's focus region. A preset number of device focus regions are acquired sequentially based on the spatial distances corresponding to each device's focus region, from smallest to largest, and these are recorded as a region allocation set for that processing unit. If a device focus region has already been allocated to the region allocation set of another processing unit, it will no longer participate in the region allocation for that processing unit. The preset allocation number is the ratio of the number of IoT sensing devices within the monitoring area to the storage capacity of the processing unit.

[0050] Please see Figure 4 As shown, this is a flowchart of an embodiment of the present invention for determining whether to perform task unloading. The dynamic processing module further includes an adaptive analysis unit, which is used to determine, based on the model demand update rate of the processing unit, to perform task unloading on the region allocation set where the model demand update rate is greater than the preset model demand update rate. The task unloading target is the IoT sensing device with the largest disturbance frequency. The task is a process in which the pre-trained model processes the uploaded sequences of each IoT sensing device within the detection period to obtain the prediction result.

[0051] The dynamic processing module also includes a demand calculation unit to obtain the model demand update rate of each processing unit. This includes determining the model demand update rate based on the number of model demand changes and the number of detection cycles within the most recent complete historical acquisition period. The model demand update rate is positively correlated with the number of model demand changes and negatively correlated with the number of detection cycles. In this embodiment, the model demand update rate is the ratio of the number of model demand changes to the number of detection cycles within the most recent complete historical acquisition period. A model demand change is triggered when the average numerical rise in the focused area of ​​each device in a single processing unit exceeds a preset average numerical rise. The average numerical rise is the average value of the monitoring rise of each IoT sensing device in the focused area of ​​a single device. For a single IoT sensing device, the monitoring rise is the slope when linearly fitting the monitoring values ​​corresponding to the data types in the uploaded sequence of the most recent complete detection period. The monitoring values ​​include, but are not limited to, the pressure values ​​of gas pipelines, the concentration values ​​of combustible gases in adjacent underground spaces, the flow rates of water supply networks, the proportion of cracked areas in images of bridge structure crack monitoring, and the proportion of waterlogged areas in videos of urban road waterlogging monitoring. Understandably, the model update rate is used to determine how frequently the pre-trained model in the processing unit needs to be re-adapted due to changes in the distribution of input data. The higher the model update rate, the higher the proportion of devices with continuously rising monitoring values ​​in the area covered by the processing unit. The changes in the monitoring values ​​may have shifted from the baseline during the model training phase, and the reliability of the results may decrease.

[0052] The dynamic processing module also includes a disturbance calculation unit to calculate the disturbance frequency of the IoT sensing device. This includes determining the disturbance frequency based on the number of detection cycles where the monitoring rise value of the IoT sensing device exceeds a preset monitoring rise value and the duration of the historical acquisition cycle. The disturbance frequency is positively correlated with the number of detection cycles exceeding the threshold and negatively correlated with the duration of the historical acquisition cycle. In this embodiment, the disturbance frequency is the ratio of the number of detection cycles where the monitoring rise value of the IoT sensing device exceeds the preset monitoring rise value to the duration of the historical acquisition cycle within the most recent complete historical acquisition cycle. It is understood that the disturbance frequency is used to determine the frequency with which a single IoT sensing device generates highly fluctuating data within a historical acquisition cycle. The higher the disturbance frequency, the longer the device's data remains in an abnormal fluctuation state deviating from the historical baseline, which can easily lead to resource constraints in the processing unit due to excessively high model update requirements.

[0053] In this embodiment of the invention, obtaining the preset model requirement update rate includes: simulating the operation of multiple sets of processing unit data with different model requirement update rates; setting different model requirement update rate thresholds for task unloading judgment tests; and selecting the average value of the model requirement update rate thresholds that satisfy user needs for the lifeline prediction rate after unloading as the preset model requirement update rate. In this embodiment, the prediction rate of the lifeline after unloading is the ratio of the total amount of standard data processed by each processing unit within a single detection cycle to the duration of that detection cycle after task unloading. The higher the prediction rate, the more timely the platform's perception and prediction of the lifeline project status. The greater the user's sensitivity to the timeliness of lifeline project prediction, the higher the requirement for the prediction rate when selecting the preset value using the prediction rate of the lifeline after unloading as an evaluation index. The higher the user's requirement for processing efficiency, the smaller the value of the preset model requirement update rate.

[0054] In this embodiment of the invention, obtaining the preset average numerical rise rate includes: simulating the process by collecting monitoring data from multiple IoT sensing devices with different average numerical rise rates; setting different average numerical rise rate thresholds for model requirement change triggering tests; and selecting the average of the average numerical rise rate thresholds that satisfy user requirements for the lifeline prediction rate after unloading as the preset average numerical rise rate. The higher the user's requirement for timely model updates, the smaller the value of the preset average numerical rise rate.

[0055] In this embodiment of the invention, obtaining the preset monitoring rise value includes: simulating the process by collecting monitoring data from multiple IoT sensing devices with different monitoring rise values; calculating and testing the disturbance frequency by setting different monitoring rise value thresholds; and selecting the minimum monitoring rise value threshold that satisfies the user's requirements for the predicted lifeline rate after unloading as the preset monitoring rise value. The higher the user's requirement for the sensitivity of disturbance data identification, the smaller the value of the preset monitoring rise value.

[0056] Specifically, the dynamic processing module also includes a benchmark analysis unit, which is used to determine whether to adjust the task unloading benchmark from task uniformity to running interval compliance when the task diversity balance of the processing unit is greater than the preset task diversity balance. It also responds to the condition that the task diversity balance of the processing unit is less than or equal to the preset task diversity balance, and determines the task unloading benchmark as task singleness.

[0057] The dynamic processing module also includes a baseline calculation unit for calculating the task diversity balance, task singularity, and operation interval compliance of the processing units. The task diversity balance is obtained by determining the task diversity balance based on the standard deviation of the number of pre-trained models called in each processing unit. The task diversity balance is negatively correlated with the standard deviation of the number of pre-trained models called in each processing unit. In this embodiment, the task diversity balance is the reciprocal of the standard deviation of the number of pre-trained models called in each processing unit. It can be understood that the task diversity balance is used to determine the degree of distribution balance of task types among the processing units. The greater the task diversity balance, the closer the number of pre-trained models called by each processing unit, and the more evenly distributed the task types undertaken by each unit. The pre-trained model is an analysis model pre-trained using historical monitoring data from various fields of urban lifeline engineering as training samples, based on a deep learning algorithm. It is used to perform risk analysis and status prediction when the processing unit calls data based on the data type of the equipment's focused area, including but not limited to gas leak risk analysis models, gas pipeline pressure prediction models, water supply pipeline burst early warning models, drainage pipeline overflow early warning models, and bridge structure damage identification models. The training methods and specific network structures of the model are known to those skilled in the art and will not be described in detail here.

[0058] The process of obtaining task singleness includes: the task singleness is determined based on the maximum number of IoT sensing devices processed by each pre-trained model in a single processing unit and the total number of IoT sensing devices in the corresponding region allocation set of that processing unit. The task singleness is positively correlated with the maximum number of devices and negatively correlated with the total number of devices in the region allocation set. In this embodiment, the task singleness is the ratio of the maximum number of IoT sensing devices processed by each pre-trained model in a single processing unit to the total number of IoT sensing devices in the corresponding region allocation set of that processing unit. It can be understood that task singleness is used to determine the concentration of task types in a single processing unit. The higher the task singleness, the more similar the types of IoT sensing devices processed by a certain type of pre-trained model in that processing unit, eliminating the need for frequent switching.

[0059] The process of obtaining the operating interval compliance includes: the operating interval compliance is determined based on the ratio of the maximum intermittent duration of the processing unit within the most recent complete historical acquisition cycle to the duration of the historical acquisition cycle. The operating interval compliance is positively correlated with the maximum intermittent duration and negatively correlated with the duration of the historical acquisition cycle. In this embodiment, for a single processing unit, the operating interval compliance is the ratio of the maximum intermittent duration of the processing unit within the most recent complete historical acquisition cycle to the duration of the historical acquisition cycle, during which the processing unit did not receive standard data from any IoT sensing devices for processing. It can be understood that the operating interval compliance is used to determine the longest continuous period of time during which the processing unit is idle within a historical cycle. The larger the operating interval compliance, the longer the continuous idle window of the processing unit, indicating a more sufficient capacity to handle additional IoT sensing devices.

[0060] In this embodiment of the invention, obtaining the preset task diversity balance includes: simulating the operation of processing units with different task diversity balances by collecting multiple sets of data; setting different task diversity balance thresholds for unloading baseline switching tests; and selecting the minimum value of the task diversity balance threshold where the predicted lifeline rate after unloading meets the user's requirements as the preset task diversity balance. The higher the user's requirement for the balance of task allocation in the processing units, the smaller the value of the preset task diversity balance.

[0061] Specifically, the dynamic processing module also includes an unloading allocation unit, which is used to migrate the unloaded IoT sensing devices to the area allocation set corresponding to the processing unit with the largest task unloading benchmark.

[0062] The technical solution of the present invention has been described above with reference to the preferred embodiments shown in the accompanying drawings. However, it will be readily understood by those skilled in the art that the scope of protection of the present invention is obviously not limited to these specific embodiments. Without departing from the principles of the present invention, those skilled in the art can make equivalent changes or substitutions to the relevant technical features, and the technical solutions after these changes or substitutions will all fall within the scope of protection of the present invention.

[0063] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A platform for detecting and aggregating data from IoT sensing devices in urban lifeline engineering projects, characterized in that, include: The protocol parsing module is used to access various IoT sensing devices in the urban lifeline project. Based on the communication protocol type of each device, it calls the corresponding protocol parser to convert the raw sensing data into standard data in a unified format. The monitoring and analysis module is used to determine whether the standard data is disturbed data based on the data integrity and time series fluctuation of the standard data. The region division module is used to determine whether to adjust the region selection benchmark from urban distribution characteristics to interference distribution characteristics based on the stability of the source distribution of disturbance data, so as to divide the equipment focusing area based on the region selection benchmark; The dynamic processing module is used to construct several regional allocation sets based on spatial distance and determine whether to unload tasks based on the model requirement update rate of the processing unit. The dynamic processing module includes several processing units, which are used to analyze and process standard data using a pre-trained model.

2. The urban lifeline engineering IoT sensing device detection and data aggregation platform according to claim 1, characterized in that, The monitoring and analysis module includes a disturbance analysis unit, which is used to identify standard data with a data integrity less than a preset data integrity or a time series fluctuation greater than a preset time series fluctuation, and determine them as disturbance data.

3. The urban lifeline engineering IoT sensing device detection and data aggregation platform according to claim 2, characterized in that, The monitoring and analysis module also includes a data analysis unit, which is used to determine the data integrity based on the quantity of key data types and to determine the time series fluctuation based on the upload time difference of adjacent standard data combinations. The data completeness is positively correlated with the number of key data types. The time-series volatility is positively correlated with the upload time difference of adjacent standard data combinations.

4. The urban lifeline engineering IoT sensing device detection and data aggregation platform according to claim 3, characterized in that, The region division module includes a statistical analysis unit and a division selection unit; The statistical analysis unit is used to determine the source distribution stability based on the spatiotemporal change rate of the IoT sensing devices corresponding to the disturbance data; The selection unit is divided to respond to the condition that the source distribution stability is greater than the preset source distribution stability, and to determine whether to adjust the regional selection benchmark from urban distribution characteristics to interference distribution characteristics.

5. The urban lifeline engineering IoT sensing device detection and data aggregation platform according to claim 4, characterized in that, The region division module also includes a division execution unit, which is used to respond to the condition that the region selection benchmark is the interference distribution characteristics, and to construct regions for the IoT sensing devices with disturbance data in descending order of stable sub-values, so as to obtain a number of device focus areas that meet the preset construction conditions. The preset construction condition is that the average value of the stable sub-values ​​of IoT sensing devices within the device's focal area is less than or equal to the preset average stable sub-value.

6. The urban lifeline engineering IoT sensing device detection and data aggregation platform according to claim 5, characterized in that, The partitioning execution unit is also used to respond to the condition that the regional selection benchmark is the urban distribution characteristics, and to construct regions for the IoT sensing devices with disturbance data in descending order of regional building density in order to obtain the device focus area.

7. The urban lifeline engineering IoT sensing device detection and data aggregation platform according to claim 1, characterized in that, The dynamic processing module includes a space allocation unit, which is used to allocate the focus area of ​​each device based on the ascending spatial distance to obtain several set of area allocations; The regional allocation set includes processing units and several device focus areas.

8. The urban lifeline engineering IoT sensing device detection and data aggregation platform according to claim 7, characterized in that, The dynamic processing module also includes an adaptive analysis unit, which is used to determine the task offloading for the region allocation set whose model demand update rate is greater than the preset model demand update rate based on the condition of the model demand update rate of the processing unit. The task offloading target is the IoT sensing device with the largest disturbance frequency.

9. The urban lifeline engineering IoT sensing device detection and data aggregation platform according to claim 8, characterized in that, The dynamic processing module also includes a benchmark analysis unit, which is used to determine to adjust the task unloading benchmark from task uniformity to running interval compliance in response to the condition that the task diversity balance of the processing unit is greater than the preset task diversity balance.

10. The urban lifeline engineering IoT sensing device detection and data aggregation platform according to claim 9, characterized in that, The dynamic processing module also includes an unloading allocation unit, which is used to migrate the unloaded IoT sensing devices to the area allocation set corresponding to the processing unit with the largest task unloading benchmark.

Images