Smart city digital monitoring system based on the Internet of Things

Abstract

The invention discloses a digital monitoring system for a smart city based on the Internet of Things, which relates to the technical field of the Internet of Things and comprises a monitoring center, wherein the monitoring center comprises a data analysis module, a clustering layout module, a real-time monitoring module and an ultra-short-term prediction module; the criticality level corresponding to each type of monitoring data of each building and infrastructure is obtained; the starting area point corresponding to each type of monitoring data is obtained, a regional clustering operation of the starting area point is performed, the aggregation area corresponding to each type of monitoring data is obtained, and sensor terminals are deployed according to the aggregation area; a time convolution prediction model is constructed, the judgment data is obtained, the monitoring results of each sensor terminal are analyzed in real time, and abnormal alarm information of each building and infrastructure is obtained; ultra-short-term prediction is performed on the building or infrastructure that generates abnormal alarm information of abnormal type monitoring data, and an ultra-short-term prediction result is obtained; and the real-time, accuracy and monitoring efficiency of the digital monitoring process of the smart city are significantly improved.

Description

Technical Field

[0001] The present invention relates to the technical field of Internet of Things, and in particular to a digital monitoring system for a smart city based on the Internet of Things. Background Art

[0002] The prior art CN115278180A "A smart city monitoring system based on the Internet of Things" includes: an aerial monitoring module, a ground monitoring module, an underground monitoring module and a personnel wearable module. The aerial monitoring module is a high-definition shooting device placed in the air, the ground monitoring module is a shooting device and a positioning device installed on the ground, and the underground monitoring module is a multi-wireless induction-based sensor pre-buried underground; the personnel wearable module is an intelligent terminal device. The present invention can realize all-round monitoring of people in the city through the aerial monitoring module, the ground monitoring module and the underground monitoring module set in conjunction with the personnel wearable module worn by the personnel, and can also judge the behavior of the personnel, and perform image recognition and video recording forensics functions.

[0003] The prior art CN118396231A "A smart city digital management system and management method" includes an urban infrastructure management module, a traffic management module, an environmental detection module, a public safety module, an urban governance module, an energy management module, a data analysis and decision support module, and an intelligent service module. The modules are integrated and work together to achieve comprehensive coverage of urban management by integrating multiple modules such as urban infrastructure management, traffic management, environmental detection, public safety, urban governance, energy management, data analysis and decision support, and intelligent services.

[0004] Existing smart city digital monitoring systems face a variety of technical and implementation challenges, including possible errors in data collection and processing, which affect the quality of decision-making, data collection and analysis of events are often not comprehensive and accurate, the hardware and software of the monitoring system may have stability issues, affecting its reliable operation, and a large amount of real-time data requires efficient processing and analysis capabilities, which places high demands on computing resources. Summary of the invention

[0005] In order to solve the above technical problems, the object of the present invention is to provide a digital monitoring system for smart cities based on the Internet of Things, including a monitoring center, wherein the monitoring center includes a data analysis module, a clustering layout module, a real-time monitoring module and an ultra-short-term prediction module;

[0006] The data analysis module is used to obtain the criticality level corresponding to each type of monitoring data of each building and infrastructure according to the functional characteristics and historical monitoring data of each building and infrastructure in the target urban area;

[0007] The clustering deployment module is used to obtain the starting area points corresponding to each type of monitoring data according to the criticality level corresponding to each type of monitoring data, perform regional clustering operations on the starting area points, obtain the aggregation areas corresponding to each type of monitoring data, and deploy sensor terminals for each building and infrastructure according to the aggregation areas corresponding to each type of monitoring data;

[0008] The real-time monitoring module is used to construct a time convolution prediction model, obtain the interpretation data of various types of monitoring data of various buildings and infrastructures, perform real-time analysis on the monitoring results of various sensor terminals, and obtain abnormal alarm information for various buildings and infrastructures based on the analysis results;

[0009] The ultra-short-term prediction module is used to perform ultra-short-term prediction on the building or infrastructure that generates abnormal alarm information of abnormal type monitoring data, and obtain ultra-short-term prediction results.

[0010] Furthermore, according to the functional characteristics and historical monitoring data of each building and infrastructure in the target urban area, the process of obtaining the criticality level corresponding to each type of monitoring data of each building and infrastructure includes:

[0011] Acquire the building and infrastructure data of the target city area, obtain the functional characteristics and location characteristics of each building and infrastructure in the target city area based on the building and infrastructure data, use data retrieval to obtain various types of monitoring data of each building and infrastructure for the functional characteristics of each building and infrastructure, use Internet technology to identify existing data sources, and obtain historical monitoring data sets corresponding to various types of monitoring data of each building and infrastructure based on the existing data sources;

[0012] According to the historical monitoring data, the average value corresponding to each type of monitoring data of each building and infrastructure is obtained, and the numerical evaluation interval corresponding to each type of monitoring data is set. The threshold point is selected within the numerical evaluation interval to divide the numerical evaluation interval into sub-intervals with different criticality levels. The average value corresponding to each type of monitoring data of each building and infrastructure is matched with the corresponding numerical evaluation interval to obtain the criticality level corresponding to each type of monitoring data of each building and infrastructure.

[0013] Further, the process of obtaining the starting area points corresponding to each type of monitoring data according to the criticality level corresponding to each type of monitoring data, performing regional clustering operation on the starting area points, and obtaining the aggregation areas corresponding to each type of monitoring data includes:

[0014] Select a certain type of monitoring data, filter out the top n buildings or infrastructures with the highest criticality corresponding to the type of monitoring data, use the top n buildings or infrastructures as the starting area points of the type of monitoring data, obtain the Euclidean distance between each building and infrastructure and the starting area point according to the location characteristics, determine the sensor type according to the type of monitoring data, and simultaneously obtain the maximum coverage distance of the sensor type, set the clustering radius of the starting area point according to the maximum coverage distance, obtain the buildings or infrastructures whose Euclidean distance to the starting area point is less than the clustering radius, determine the level difference between the criticality level of the buildings or infrastructures and the criticality level of the starting area point, determine whether there are buildings or infrastructures whose level difference with the starting area point is less than a preset level difference, and if not, end the regional clustering operation of the starting area point, and perform a regional classification coverage range determination operation;

[0015] If any, the building or infrastructure whose grade difference with the starting area point is less than the preset grade difference will be marked as a branch area point. The branch area point will repeat the regional clustering operation, and the clustering radius of the branch area point will be set according to the maximum coverage distance. The buildings or infrastructure within the clustering radius of the branch area point whose grade difference with the starting area point is less than the preset grade difference will be marked as branch area points, and so on, until there is no building or infrastructure whose grade difference with the starting area point is less than the preset grade difference within the clustering radius of the branch area point, the regional clustering operation of the starting area point is terminated, and the regional classification coverage range determination operation is performed.

[0016] Furthermore, the process of determining the regional classification coverage includes:

[0017] When the regional clustering operation of the starting area point corresponding to the type monitoring data is completed, the aggregation area corresponding to the type monitoring data is constructed according to the starting area point and the branch area point, and at the same time, it is determined whether there are buildings or infrastructures in the target city area that are not included in the aggregation area corresponding to the type monitoring data. If so, the top k buildings or infrastructures with the highest criticality level corresponding to the type monitoring data are screened out from the buildings or infrastructures not included in the aggregation area, and the top k buildings or infrastructures are used as the starting area points for regional clustering operation. When the regional clustering operation of the starting area point is completed, the aggregation area of ​​the starting area point is constructed, and then it is determined whether there are buildings or infrastructures in the target city area that are not included in the aggregation area corresponding to the type monitoring data. If so, the above regional clustering operation is repeated until there are no buildings or infrastructures in the target city area that are not included in the aggregation area corresponding to the type monitoring data, and the aggregation areas of several starting area points corresponding to the type monitoring data are obtained.

[0018] Furthermore, the process of obtaining the aggregation areas corresponding to each type of monitoring data in the target urban area and deploying sensor terminals for each building and infrastructure includes:

[0019] A monitoring indicator comparison table is constructed, wherein the monitoring indicator comparison table includes monitoring frequencies and sensor deployment quantities corresponding to different criticality levels. According to the criticality level of the starting area point of the aggregation area of ​​several starting area points corresponding to each type of monitoring data and the monitoring indicator comparison table, the deployment quantity and monitoring frequency of each type of sensor of each building and infrastructure are obtained. According to the deployment quantity and monitoring frequency of each type of sensor, sensor terminals of each building and infrastructure are deployed. The sensor terminals are used to monitor each type of monitoring data of each building and infrastructure, mark the monitoring time, and set the monitoring cycle.

[0020] Furthermore, the process of constructing a temporal convolution prediction model and obtaining interpretation data of various types of monitoring data of various buildings and infrastructures includes:

[0021] A time convolution prediction model is constructed based on deep learning, and the historical monitoring data sets corresponding to various types of monitoring data of various buildings and infrastructure are used as training sets and test sets. The training sets are input into the time convolution prediction model for training until the loss function training is stable, and the model parameters are saved. The time convolution prediction model is tested by the test set until it meets the preset requirements, and the time convolution prediction model is output;

[0022] According to the time convolution prediction model, the predicted numerical time series sequence corresponding to each type of monitoring data of each building and infrastructure in the current monitoring period is output, and the predicted numerical time series sequence corresponding to each type of monitoring data of each building and infrastructure is marked as judgment data.

[0023] Furthermore, the monitoring results of each sensor terminal are analyzed in real time, and the process of obtaining abnormal alarm information of each building and infrastructure according to the analysis results includes:

[0024] Obtain the real-time numerical time series sequence corresponding to each type of monitoring data of each sensor terminal of each building and infrastructure in the target area, compare the real-time numerical time series sequence corresponding to each type of monitoring data with the corresponding judgment data, obtain the deviation rate information corresponding to each type of monitoring data, obtain the cumulative time when the deviation rate information corresponding to each type of monitoring data is greater than a preset deviation rate threshold, and mark the type of monitoring data with a cumulative time greater than the cumulative time threshold as abnormal type monitoring data;

[0025] Acquire the location features of the sensor terminal that collects the abnormal type monitoring data and other sensor terminals, acquire the Euclidean distance between other sensor terminals and the sensor terminal and the maximum coverage distance of the sensor terminal, screen out other sensor terminals whose Euclidean distance to the sensor terminal is less than the maximum coverage distance, and simultaneously acquire the monitoring data difference corresponding to each type of monitoring data between the other sensor terminals and the sensor terminal according to the judgment data, set the difference threshold according to the monitoring data difference, acquire several monitoring data difference values ​​about the abnormal type monitoring data between the sensor terminal and other sensor terminals, and compare the monitoring data difference with the difference threshold;

[0026] If the differences of several monitoring data are all less than or equal to the difference threshold, abnormal alarm information of abnormal type monitoring data is generated; if there is a monitoring data difference greater than the difference threshold, abnormal alarm information of the sensor terminal of abnormal type monitoring data is generated.

[0027] Furthermore, an ultra-short-term prediction is performed on the building or infrastructure that generates abnormal alarm information of abnormal type monitoring data. The process of obtaining the ultra-short-term prediction result includes:

[0028] Construct an ultra-short-term prediction rule table, the ultra-short-term prediction rule table includes prediction durations and monitoring frequencies corresponding to different deviation rate information, when generating abnormal alarm information of abnormal type monitoring data, obtain the deviation rate information corresponding to the type monitoring data, and obtain the monitoring frequency of the sensor terminal of the abnormal type monitoring data and the prediction duration of the time convolution prediction model according to the deviation rate information and the ultra-short-term prediction rule table;

[0029] According to the monitoring frequency of the sensor terminal of the abnormal type monitoring data, the monitoring frequency of the sensor terminal in the current collection period is adjusted, and the historical monitoring data sets corresponding to the various types of monitoring data of the building or infrastructure where the sensor terminal is located and the various types of monitoring data collected by the sensor terminal in the current monitoring period are used as training data. The training data is used to train the time convolution prediction model, and the trained time convolution prediction model outputs the predicted numerical time series sequence of each type of monitoring data within the prediction duration.

[0030] Compared with the prior art, the present invention has the following beneficial effects:

[0031] 1. The data analysis module is responsible for collecting and analyzing historical monitoring data from different buildings and infrastructures, determining the importance (criticality level) of these data, and through in-depth analysis of historical data, identifying which data types are most critical for predicting potential problems. It can automatically filter out important data types, reduce unnecessary data processing burden, and improve efficiency.

[0032] 2. Clustering deployment module Based on the criticality level given by the data analysis module, this module decides in which areas to deploy more sensors to ensure that important facilities are more intensively monitored, optimizes the layout of sensors, ensures that resources are used efficiently, and covers key areas to improve the targeted nature of monitoring.

[0033] 3. The real-time monitoring module builds a time convolution prediction model, which can process the data received from various sensors in real time, detect abnormal situations immediately, and trigger alarms, thereby improving the response speed, being able to respond to emergencies quickly, and reducing losses.

[0034] 4. Ultra-short-term prediction module For situations where alarms have been triggered, this module will further analyze the data and make ultra-short-term predictions to help decision makers better understand the development trend of events, provide deeper information support, and help formulate timely and effective response measures. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] Figure 1 This is a schematic diagram of a smart city digital monitoring system based on the Internet of Things according to an embodiment of the present application. DETAILED DESCRIPTION

[0036] The following is a clear and complete description of the technical solutions in the embodiments of the present application in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of the present application.

[0037] like Figure 1 As shown, the digital monitoring system of a smart city based on the Internet of Things includes a monitoring center, which includes a data analysis module, a clustering layout module, a real-time monitoring module and an ultra-short-term prediction module;

[0038] The data analysis module is used to obtain the criticality level corresponding to each type of monitoring data of each building and infrastructure according to the functional characteristics and historical monitoring data of each building and infrastructure in the target urban area;

[0039] The clustering deployment module is used to obtain the starting area points corresponding to each type of monitoring data according to the criticality level corresponding to each type of monitoring data, perform regional clustering operations on the starting area points, obtain the aggregation areas corresponding to each type of monitoring data, and deploy sensor terminals for each building and infrastructure according to the aggregation areas corresponding to each type of monitoring data;

[0040] The real-time monitoring module is used to construct a time convolution prediction model, obtain the interpretation data of various types of monitoring data of various buildings and infrastructures, perform real-time analysis on the monitoring results of various sensor terminals, and obtain abnormal alarm information for various buildings and infrastructures based on the analysis results;

[0041] The ultra-short-term prediction module is used to perform ultra-short-term prediction on the building or infrastructure that generates abnormal alarm information of abnormal type monitoring data, and obtain ultra-short-term prediction results.

[0042] It should be further explained that, in the specific implementation process, according to the functional characteristics of each building and infrastructure in the target urban area and the historical monitoring data, the process of obtaining the criticality level corresponding to each type of monitoring data of each building and infrastructure includes:

[0043] Acquire the building and infrastructure data of the target urban area, acquire the functional characteristics and location characteristics of each building and infrastructure in the target urban area based on the building and infrastructure data, and use data retrieval to acquire various types of monitoring data of each building and infrastructure for the functional characteristics of each building and infrastructure, wherein the types of monitoring data include traffic flow data, air quality data, public safety data and energy consumption data, wherein the air quality data includes the concentration of various pollutants, such as sulfur dioxide, nitrogen dioxide, carbon monoxide, ozone, PM2.5 (particles with a diameter of less than or equal to 2.5 microns), PM10 (particles with a diameter of less than or equal to 10 microns), etc.; public safety data includes traffic accident records (including vehicle collisions, pedestrian injuries, etc.), fire alarm records, crowd density and natural disaster warning information (such as earthquakes, floods, etc.); energy consumption data includes the use of electricity, water and other energy in buildings or infrastructure, such as electricity consumption, such as total electricity consumption, peak electricity consumption, etc., water consumption, natural gas consumption, etc.

[0044] Use Internet technology to identify existing data sources, and obtain historical monitoring data sets corresponding to various types of monitoring data for various buildings and infrastructures based on existing data sources, such as:

[0045] Obtain traffic flow data and traffic accident records through the official website of the Traffic Management Bureau, which will publish real-time or historical traffic flow data and traffic accident records on its official website, or obtain traffic flow data and traffic accident records through third-party traffic information service providers, such as Baidu Maps and Amap, which provide real-time traffic flow information and traffic accident records through cooperation with the traffic management department;

[0046] Obtain air quality data through the official website of the Environmental Protection Bureau. The Environmental Protection Bureau will publish the Air Quality Index (AQI) and related pollutant data on its official website. Or obtain air quality data through air quality monitoring stations. Air quality monitoring stations set up by local environmental protection departments will regularly publish monitoring data.

[0047] Obtain fire alarm records through the fire department, which is responsible for processing and recording fire alarms. Obtain energy consumption data of various buildings and infrastructure through power companies, water bureaus, gas companies, etc. Power companies will publish power consumption data, water bureaus will publish data on water resource consumption, including water supply and drainage, and gas companies will also publish gas consumption and other related data;

[0048] At the same time, the data provided by the above data sources are open, transparent, legal and compliant;

[0049] According to the historical monitoring data, the average value corresponding to each type of monitoring data of each building and infrastructure is obtained, and the numerical evaluation interval corresponding to each type of monitoring data is set. The threshold point is selected within the numerical evaluation interval to divide the numerical evaluation interval into sub-intervals with different criticality levels. The average value corresponding to each type of monitoring data of each building and infrastructure is matched with the corresponding numerical evaluation interval to obtain the criticality level corresponding to each type of monitoring data of each building and infrastructure.

[0050] It should be further explained that, in the specific implementation process, the starting area points corresponding to each type of monitoring data are obtained according to the criticality level corresponding to each type of monitoring data, and the regional clustering operation of the starting area points is performed. The process of obtaining the aggregation area corresponding to each type of monitoring data includes:

[0051] Select a certain type of monitoring data corresponding to each building and infrastructure, and screen out the top n buildings or infrastructures with the highest criticality level corresponding to the type of monitoring data, where n is determined by the total number of buildings and infrastructures and the selection ratio, and usually the number of n is half of the total number of buildings and infrastructures, and use the top n buildings or infrastructures as the starting area points of the type monitoring data, obtain the Euclidean distance between each building and infrastructure and the starting area point according to the location characteristics of each building and infrastructure, determine the sensor type according to the type monitoring data, and simultaneously obtain the maximum coverage distance of the sensor type, set the clustering radius of the starting area point according to the maximum coverage distance, obtain the building or infrastructure whose Euclidean distance to the starting area point is less than the clustering radius, determine the level difference between the criticality level of the building or infrastructure and the criticality level of the starting area point, and determine whether there is a building or infrastructure whose level difference with the starting area point is less than a preset level difference. If not, end the regional clustering operation of the starting area point, and perform a regional classification coverage range determination operation;

[0052] If any, the building or infrastructure whose level difference with the starting area point is less than the preset level difference will be marked as a branch area point, and the branch area point will repeat the regional clustering operation, and the clustering radius of the branch area point will be set according to the maximum coverage distance, and the building or infrastructure whose Euclidean distance with the branch area point is less than the clustering radius will be obtained, and the level difference between the building or infrastructure and the branch area point will be obtained, and the building or infrastructure whose level difference is less than the level difference threshold will be marked as a branch area point, and so on, until there is no building or infrastructure whose level difference with the starting area point is less than the preset level difference within the clustering radius of the branch area point, the regional clustering operation of the starting area point is terminated, and the regional classification coverage range determination operation is performed.

[0053] It should be further explained that, in the specific implementation process, the process of determining the regional classification coverage includes:

[0054] When the regional clustering operation of the starting area point corresponding to the type monitoring data is completed, the aggregation area corresponding to the type monitoring data is constructed according to the starting area point and the branch area point, and at the same time, it is determined whether there are buildings or infrastructures in the target city area that are not included in the aggregation area corresponding to the type monitoring data. If so, the top k buildings or infrastructures with the highest criticality level corresponding to the type monitoring data are screened out from the buildings or infrastructures not included in the aggregation area, wherein k is determined by the total number of buildings and infrastructures not included in the starting aggregation area and the selection ratio. Usually, the number n is half of the total number of buildings and infrastructures not included in the starting aggregation area. The top k buildings or infrastructures are used as the starting area points for regional clustering operation. When the regional clustering operation of the starting area point is completed, the aggregation area of ​​the starting area point is constructed, and then it is determined whether there are buildings or infrastructures in the target city area that are not included in the aggregation area corresponding to the type monitoring data. If so, the above regional clustering operation is repeated until there are no buildings or infrastructures in the target city area that are not included in the aggregation area corresponding to the type monitoring data, and the aggregation areas of several starting area points corresponding to the type monitoring data are obtained.

[0055] It should be further explained that, in the specific implementation process, the process of obtaining the aggregation area corresponding to each type of monitoring data in the target urban area and deploying sensor terminals for each building and infrastructure includes:

[0056] A monitoring indicator comparison table is constructed, wherein the monitoring frequency and sensor deployment quantity corresponding to different criticality levels are included; the deployment quantity and monitoring frequency of each type of sensor of each building and infrastructure are obtained according to the criticality level of the starting area point of the aggregation area of ​​several starting area points corresponding to each type of monitoring data and the monitoring indicator comparison table; sensor terminals are deployed for each building and infrastructure according to the deployment quantity and monitoring frequency of each type of sensor; the sensor terminals are used to monitor each type of monitoring data of each building and infrastructure and mark the monitoring time; and a monitoring period is set, and the length of the monitoring period is usually one day.

[0057] It should be further explained that, in the specific implementation process, the process of constructing a time convolution prediction model and obtaining interpretation data of various types of monitoring data of various buildings and infrastructures includes:

[0058] A time convolution prediction model is constructed based on deep learning, and the historical monitoring data sets corresponding to various types of monitoring data of various buildings and infrastructure are used as training sets and test sets. The training sets are input into the time convolution prediction model for training until the loss function training is stable, and the model parameters are saved. The time convolution prediction model is tested by the test set until it meets the preset requirements, and the time convolution prediction model is output;

[0059] According to the time convolution prediction model, the predicted numerical time series sequence corresponding to each type of monitoring data of each building and infrastructure in the current monitoring period is output, and the predicted numerical time series sequence corresponding to each type of monitoring data of each building and infrastructure is marked as judgment data.

[0060] It should be further explained that, in the specific implementation process, the monitoring results of each sensor terminal are analyzed in real time, and the process of obtaining abnormal alarm information of each building and infrastructure according to the analysis results includes:

[0061] Obtain the real-time numerical time series sequence corresponding to each type of monitoring data of each sensor terminal of each building and infrastructure in the target area, compare the real-time numerical time series sequence corresponding to each type of monitoring data with the corresponding judgment data, obtain the deviation rate information corresponding to each type of monitoring data, obtain the cumulative time when the deviation rate information corresponding to each type of monitoring data is greater than a preset deviation rate threshold, and mark the type of monitoring data with a cumulative time greater than the cumulative time threshold as abnormal type monitoring data;

[0062] Acquire the location features of the sensor terminal that collects the abnormal type monitoring data and other sensor terminals, acquire the Euclidean distance between other sensor terminals and the sensor terminal and the maximum coverage distance of the sensor terminal, screen out other sensor terminals whose Euclidean distance to the sensor terminal is less than the maximum coverage distance, and simultaneously acquire the monitoring data difference corresponding to each type of monitoring data between the other sensor terminals and the sensor terminal according to the judgment data, set the difference threshold according to the monitoring data difference, acquire several monitoring data difference values ​​about the abnormal type monitoring data between the sensor terminal and other sensor terminals, and compare the monitoring data difference with the difference threshold;

[0063] The calculation formula for obtaining the difference in monitoring data corresponding to each type of monitoring data between the other sensor terminal and the sensor terminal is:

[0064] ;

[0065] in, represents the difference in monitoring data between the sensor terminal and other sensor terminals regarding the i-th type of monitoring data; represents the predicted data of the i-th type of monitoring data of the sensor terminal at the t-th time; It represents the predicted data of the i-th type of monitoring data of other sensor terminals at the t-th time; n represents the total number of time points in the monitoring cycle.

[0066] If the differences of several monitoring data are all less than or equal to the difference threshold, abnormal alarm information of abnormal type monitoring data is generated; if there is a monitoring data difference greater than the difference threshold, abnormal alarm information of the sensor terminal of abnormal type monitoring data is generated.

[0067] It should be further explained that, in the specific implementation process, the process of performing ultra-short-term prediction on the building or infrastructure that generates abnormal alarm information of abnormal type monitoring data and obtaining the ultra-short-term prediction result includes:

[0068] Construct an ultra-short-term prediction rule table, the ultra-short-term prediction rule table includes prediction durations and monitoring frequencies corresponding to different deviation rate information, when generating abnormal alarm information of abnormal type monitoring data, obtain the deviation rate information corresponding to the type monitoring data, and obtain the monitoring frequency of the sensor terminal of the abnormal type monitoring data and the prediction duration of the time convolution prediction model according to the deviation rate information and the ultra-short-term prediction rule table;

[0069] The monitoring frequency of the sensor terminal in the current collection cycle is adjusted according to the monitoring frequency of the sensor terminal of the abnormal type monitoring data, and the historical monitoring data sets corresponding to the various types of monitoring data of the building or infrastructure where the sensor terminal is located and the various types of monitoring data collected by the sensor terminal in the current monitoring cycle are used as training data. The training data is used to train the time convolution prediction model, and the trained time convolution prediction model outputs the predicted numerical time series sequence of each type of monitoring data within the prediction duration, and the predicted numerical time series sequence of each type of monitoring data is used to provide a basis for emergency measures.

Claims

1. The digital monitoring system of smart city based on Internet of Things is characterized by: It includes a monitoring center, which includes a data analysis module, a clustering layout module, a real-time monitoring module and an ultra-short-term prediction module; The data analysis module is used to obtain the criticality level corresponding to each type of monitoring data of each building and infrastructure according to the functional characteristics and historical monitoring data of each building and infrastructure in the target urban area, including: Acquire the building and infrastructure data of the target city area, obtain the functional characteristics and location characteristics of each building and infrastructure in the target city area based on the building and infrastructure data, use data retrieval to obtain various types of monitoring data of each building and infrastructure for the functional characteristics of each building and infrastructure, use Internet technology to identify existing data sources, and obtain historical monitoring data sets corresponding to various types of monitoring data of each building and infrastructure based on the existing data sources; According to the historical monitoring data set, the average value corresponding to each type of monitoring data of each building and infrastructure is obtained, and at the same time, the numerical evaluation interval corresponding to each type of monitoring data is set, and a threshold point is selected in the numerical evaluation interval to divide the numerical evaluation interval into sub-intervals with different criticality levels, and the criticality level corresponding to each type of monitoring data of each building and infrastructure is obtained according to the average value corresponding to each type of monitoring data of each building and infrastructure and the corresponding numerical evaluation interval. The clustering deployment module is used to obtain the starting area points corresponding to each type of monitoring data according to the criticality level corresponding to each type of monitoring data, perform regional clustering operations on the starting area points, obtain the aggregation areas corresponding to each type of monitoring data, and deploy sensor terminals for each building and infrastructure according to the aggregation areas corresponding to each type of monitoring data; According to the criticality level corresponding to each type of monitoring data, the starting area point corresponding to each type of monitoring data is obtained, and the regional clustering operation of the starting area point is performed. The process of obtaining the aggregation area corresponding to each type of monitoring data includes: Select a certain type of monitoring data, filter out the top n buildings or infrastructures with the highest criticality corresponding to the type of monitoring data, use the top n buildings or infrastructures as the starting area points of the type of monitoring data, obtain the Euclidean distance between each building and infrastructure and the starting area point according to the location characteristics, determine the sensor type according to the type of monitoring data, and simultaneously obtain the maximum coverage distance of the sensor type, set the clustering radius of the starting area point according to the maximum coverage distance, obtain the buildings or infrastructures whose Euclidean distance to the starting area point is less than the clustering radius, determine the level difference between the criticality level of the buildings or infrastructures and the criticality level of the starting area point, determine whether there are buildings or infrastructures whose level difference with the starting area point is less than a preset level difference, and if not, end the regional clustering operation of the starting area point, and perform a regional classification coverage range determination operation; If there is, the building or infrastructure whose grade difference with the starting area point is less than the preset grade difference is marked as a branch area point, and the branch area point repeats the regional clustering operation, sets the clustering radius of the branch area point according to the maximum coverage distance, and marks the building or infrastructure whose grade difference with the starting area point is less than the preset grade difference within the clustering radius of the branch area point as a branch area point, and so on, until there is no building or infrastructure whose grade difference with the starting area point is less than the preset grade difference within the clustering radius of the branch area point, then the regional clustering operation of the starting area point is terminated, and the regional classification coverage range determination operation is performed; The process of determining the coverage of regional classification includes: When the regional clustering operation of the starting area point corresponding to the type monitoring data is completed, the aggregation area corresponding to the type monitoring data is constructed according to the starting area point and the branch area point, and at the same time, it is determined whether there are buildings or infrastructures in the target city area that are not included in the aggregation area corresponding to the type monitoring data. If so, the top k buildings or infrastructures with the highest criticality level corresponding to the type monitoring data are screened out from the buildings or infrastructures not included in the aggregation area, and the top k buildings or infrastructures are used as the starting area points for regional clustering operation. When the regional clustering operation of the starting area point is completed, the aggregation area of ​​the starting area point is constructed, and then it is determined whether there are buildings or infrastructures in the target city area that are not included in the aggregation area corresponding to the type monitoring data. If so, the above regional clustering operation is repeated until there are no buildings or infrastructures in the target city area that are not included in the aggregation area corresponding to the type monitoring data, and the aggregation areas of several starting area points corresponding to the type monitoring data are obtained; The process of obtaining the aggregation area corresponding to each type of monitoring data in the target urban area and deploying sensor terminals for each building and infrastructure includes: Construct a monitoring index comparison table, the monitoring index comparison table includes monitoring frequencies and sensor deployment quantities corresponding to different criticality levels, obtain the deployment quantity and monitoring frequency of each type of sensor for each building and infrastructure according to the criticality level of the starting area point of the aggregation area of ​​several starting area points corresponding to each type of monitoring data and the monitoring index comparison table, deploy sensor terminals for each building and infrastructure according to the deployment quantity and monitoring frequency of each type of sensor, the sensor terminals are used to monitor each type of monitoring data of each building and infrastructure, mark the monitoring time, and set the monitoring cycle; The real-time monitoring module is used to construct a time convolution prediction model, obtain the interpretation data of various types of monitoring data of various buildings and infrastructures, perform real-time analysis on the monitoring results of various sensor terminals, and obtain abnormal alarm information for various buildings and infrastructures based on the analysis results; The process of building a temporal convolution prediction model to obtain interpretation data of various types of monitoring data for various buildings and infrastructure includes: A time convolution prediction model is constructed based on deep learning, and the historical monitoring data sets corresponding to various types of monitoring data of various buildings and infrastructure are used as training sets and test sets. The training sets are input into the time convolution prediction model for training until the loss function training is stable, and the model parameters are saved. The time convolution prediction model is tested by the test set until it meets the preset requirements, and the time convolution prediction model is output; Output the predicted numerical time series sequence corresponding to each type of monitoring data of each building and infrastructure in the current monitoring period according to the time convolution prediction model, and mark the predicted numerical time series sequence corresponding to each type of monitoring data of each building and infrastructure as judgment data; The process of real-time analysis of the monitoring results of each sensor terminal and obtaining abnormal alarm information of each building and infrastructure based on the analysis results includes: Obtain the real-time numerical time series sequence corresponding to each type of monitoring data of each sensor terminal of each building and infrastructure in the target area, compare the real-time numerical time series sequence corresponding to each type of monitoring data with the corresponding judgment data, obtain the deviation rate information corresponding to each type of monitoring data, obtain the cumulative time when the deviation rate information corresponding to each type of monitoring data is greater than a preset deviation rate threshold, and mark the type of monitoring data with a cumulative time greater than the cumulative time threshold as abnormal type monitoring data; Acquire the location features of the sensor terminal that collects the abnormal type monitoring data and other sensor terminals, acquire the Euclidean distance between other sensor terminals and the sensor terminal and the maximum coverage distance of the sensor terminal, screen out other sensor terminals whose Euclidean distance to the sensor terminal is less than the maximum coverage distance, and simultaneously acquire the monitoring data difference corresponding to each type of monitoring data between the other sensor terminals and the sensor terminal according to the judgment data, set the difference threshold according to the monitoring data difference, acquire several monitoring data difference values ​​about the abnormal type monitoring data between the sensor terminal and other sensor terminals, and compare the monitoring data difference with the difference threshold; If the differences of several monitoring data are all less than or equal to the difference threshold, abnormal alarm information of abnormal type monitoring data is generated; if there is a monitoring data difference greater than the difference threshold, abnormal alarm information of the sensor terminal of abnormal type monitoring data is generated; The ultra-short-term prediction module is used to perform ultra-short-term prediction on the building or infrastructure that generates abnormal alarm information of abnormal type monitoring data, and obtain ultra-short-term prediction results, including: Construct an ultra-short-term prediction rule table, the ultra-short-term prediction rule table includes prediction durations and monitoring frequencies corresponding to different deviation rate information, when generating abnormal alarm information of abnormal type monitoring data, obtain the deviation rate information corresponding to the type monitoring data, and obtain the monitoring frequency of the sensor terminal of the abnormal type monitoring data and the prediction duration of the time convolution prediction model according to the deviation rate information and the ultra-short-term prediction rule table; According to the monitoring frequency of the sensor terminal of the abnormal type monitoring data, the monitoring frequency of the sensor terminal in the current collection period is adjusted, and the historical monitoring data sets corresponding to the various types of monitoring data of the building or infrastructure where the sensor terminal is located and the various types of monitoring data collected by the sensor terminal in the current monitoring period are used as training data. The training data is used to train the time convolution prediction model, and the trained time convolution prediction model outputs the predicted numerical time series sequence of each type of monitoring data within the prediction duration.

Images