Internet of Things data monitoring method and system based on artificial intelligence model analysis

Abstract

The invention discloses an Internet of Things data monitoring method and system based on artificial intelligence model analysis, and relates to the technical field of Internet of Things data analysis, and the method comprises the steps: determining the position of a data collection node, analyzing the relative relation, associating the nodes of the same function space, mapping the nodes to a virtual function space, connecting the nodes, and constructing an Internet of Things data space representation model. According to the historical exception records, extracting expressions of the model during exception and before exception, and combining the expressions into an expression group during exception; s300, performing linkage and parameter change clustering on the abnormal expressions, and classifying the abnormal expressions into a plurality of abnormal expression sets; and analyzing the performance characteristics before the abnormity, constructing a characteristic identification model, and monitoring the abnormity in real time. According to the technical scheme provided by the invention, by identifying the pre-exception performance in advance, the accuracy and real-time performance of exception early warning are improved, and the equipment risk is effectively reduced.

Description

Technical Field

[0001] The present invention relates to the technical field of Internet of Things data analysis, and in particular to an Internet of Things data monitoring method and system based on artificial intelligence model analysis. Background Art

[0002] With the rapid development of Internet of Things technology, various types of massive data are generated by data acquisition nodes and widely applied in fields such as industrial monitoring, smart home, and urban management. However, traditional Internet of Things data monitoring methods face many challenges. First, the spatial distribution of data acquisition nodes is complex, and the relative positions and functional correlations between nodes are not fully explored, resulting in a lack of integrity in data analysis. Second, existing monitoring technologies mainly focus on detection and response after anomalies occur, and it is difficult to give early warnings through data patterns before anomalies, which limits the accident prevention ability of the system. In addition, the abnormal manifestations in Internet of Things data are diverse and often accompanied by chain effects of parameter changes, and traditional methods are insufficient in abnormal classification and prediction accuracy. In recent years, the application of artificial intelligence technology in the analysis field has provided new ideas for solving the above problems, but how to combine spatial modeling, historical data feature extraction, and real-time monitoring remains a technical problem. Summary of the Invention

[0003] The purpose of the present invention is to provide a monitoring method and system that can effectively monitor anomalies in Internet of Things data.

[0004] The present invention discloses an Internet of Things data monitoring method based on artificial intelligence model analysis, including: Step S100, determining the positions of data acquisition nodes corresponding to Internet of Things data, analyzing the relative positions of different data acquisition nodes, associating data acquisition nodes belonging to the same functional space with each other, and configuring the method of collecting mapping points in a preset virtual functional space, and connecting the virtual functional spaces according to the relative position relationship to obtain an Internet of Things data space representation model; Step S200, based on historical device anomaly records, determining the manifestations during anomalies and the manifestations before anomalies of the Internet of Things data space representation model, and recording the combination of the two as the anomaly manifestation group during anomalies; Step S300, performing anomaly chain manifestation clustering and parameter change equivalent clustering on the manifestations during anomalies in sequence, and classifying the anomaly manifestation group during anomalies based on the final clustering result to obtain several anomaly manifestation sets during anomalies; Step S400, analyzing the manifestations before anomalies in the anomaly manifestation group during anomalies belonging to the same anomaly manifestation set during anomalies, constructing a feature recognition model, and using the feature recognition model to perform prior anomaly monitoring on real-time Internet of Things data.

[0005] In some embodiments disclosed by the present invention, the method for constructing an Internet of Things data space representation model includes: Step S101, calculate the average coordinate of the mapping points of the virtual function space, identify the average coordinate of the mapping points as the center point of the function space of the virtual function space, determine the distance between each acquisition mapping point and the center point of the function space, and construct a mapping connection line starting from the center point of the function space based on the distance between the acquisition mapping point and the center point of the function space and the relative direction, and set a parameter status display module at the other end of the mapping connection line; Step S102, determine the change of the corresponding parameter status display module based on the parameter change of the data acquisition point.

[0006] In some embodiments disclosed by the present invention, the method for determining the change of the parameter status display model includes: Step S1021, construct display spheres of several diameters, and each display sphere of a diameter corresponds to a preset parameter interval; Among them, the method for determining the parameter interval includes: Step S1022, perform normal change clustering, abnormal change clustering, and fault change clustering on the historical data change records of each data acquisition point, and obtain a normal data change group, an abnormal data change group, and a fault data change group respectively; Step S1023, perform an average calculation on the normal data changes in the normal data change group to obtain an average normal data change representing the normal data changes, perform an average calculation on the abnormal data changes in the abnormal data change group to obtain an average abnormal data change representing the abnormal data changes, perform an average calculation on the fault data change group to obtain an average fault data change, align the average abnormal data change, the average normal data change, and the average fault data change, and determine the average abnormal parameter, the average normal parameter, and the average fault parameter at different time nodes based on a preset reference time series, and form an average abnormal parameter sequence, an average normal parameter sequence, and an average fault parameter sequence; Step S1024, delimit several initial parameter intervals for each parameter sequence, determine the parameter quantity density mapped by the parameter sequence to each initial parameter interval, if the parameter quantity density is greater than or equal to a preset value, mark the initial parameter interval as a parameter interval to be concerned, and if there are greater than or equal to a preset number of parameter intervals to be concerned in several consecutive initial parameter intervals, fuse the heads and tails of the several consecutive initial parameter intervals to obtain a new parameter interval.

[0007] In some embodiments disclosed by the present invention, the method for determining the abnormal performance and the pre-abnormal performance of the Internet of Things data space display model includes: Step S201: Based on historical device anomaly records, determine a number of historical anomaly parameter sequences, determine the historical normal parameter sequences before each historical anomaly parameter sequence, and drive the corresponding parameter state performance module based on the corresponding historical normal parameter sequences and historical anomaly parameter sequences to obtain the abnormal performance and pre-abnormal performance of the Internet of Things data space performance model.

[0008] In some embodiments disclosed in the present invention, the method for sequentially performing abnormal chain clustering and parameter change equivalence clustering on the abnormal performance in the abnormal performance group includes: Step S301: Determine the abnormal performance of each parameter state performance module when an anomaly occurs, and determine the abnormal chain performance of the parameter state performance module based on the relative distance between the parameter state performance modules and the difference in occurrence time of the abnormal performance. The chain performance includes the parameter state performance module with abnormal performance occurring in a chain, the sequence of abnormal performance, and the difference in occurrence time; Step S302: Based on the abnormal chain performance of different parameter state performance modules in the Internet of Things data space performance model, perform abnormal chain clustering on the abnormal performance of the Internet of Things data space performance model to obtain an initial clustering set. Compare the abnormal performance of each Internet of Things data space performance model in the initial clustering set with each other, including comparing the abnormal performance of each parameter state performance module, and determine the equivalent parameter of the abnormal performance between each pair of relative parameter state performance modules. If the equivalent parameter of the abnormal performance is greater than or equal to the preset value, it is determined that the parameter changes are equivalent between the corresponding parameter state performance modules. If all pairs of relative parameter state performance modules are determined to have equivalent parameter changes, then reclassify the abnormal performance of the Internet of Things data space performance model to obtain a reference clustering set.

[0009] In some embodiments disclosed in the present invention, the method for performing abnormal chain clustering on the abnormal performance of the Internet of Things data space performance model further includes: Step S3011: Based on the relative distance between the parameter state performance modules and the difference in occurrence time of the abnormal performance, determine the abnormal chain parameter between the abnormal performances of the parameter state performance modules. Based on the abnormal chain parameter, determine whether there is an abnormal chain relationship between the parameter state performance modules, and record the combination of parameter state performance modules with an abnormal chain relationship as a chain relationship group; Step S3012: Compare the chain relationship groups between the Internet of Things data space performance models with each other. If the abnormal chain order of the parameter state performance modules between the chain relationship groups is the same, it is determined that the corresponding chain relationship groups are equivalent to each other. If all the chain relationship groups are equivalent, then classify the abnormal performance of the Internet of Things data space performance model into one category; Among them, the expression of the abnormal chain parameter between the abnormal performances of the calculation parameter status performance module is: ; Among them, is the abnormal chain parameter, is the relative distance influence adjustment coefficient, is the preset maximum relative distance, is the relative distance between parameter status performance modules, is the occurrence time difference influence adjustment coefficient, is the preset maximum occurrence time difference, is the occurrence time difference of the abnormal performance of the parameter status performance module, is the abnormal chain parameter adjustment constant.

[0010] In some embodiments disclosed by the present invention, the method for determining the equivalent parameter of abnormal performance between each relative parameter status performance module includes: Step S3021, align the abnormal performances of the parameter status performance module, and determine the performance difference amount at different time nodes according to the preset time series relationship. If the performance difference amount is less than or equal to the preset value, it is determined that the corresponding time node is an equivalent performance time node; Step S3022, determine the equivalent parameter of abnormal performance based on the proportion of equivalent time nodes of the equivalent performance time nodes occupying all time nodes.

[0011] In some embodiments disclosed by the present invention, the method for performing prior abnormal monitoring on real-time Internet of Things data by using a feature recognition model includes: Step S401, map the real-time Internet of Things data to the Internet of Things data space performance model in real time, obtain the real-time performance of the Internet of Things data space performance model, and perform feature extraction on the real-time performance to obtain a real-time performance feature sequence; Step S402, compare the real-time performance feature sequence with the abnormal pre-performance feature sequences of different abnormal pre-performances, and based on the comparison result, determine the matching abnormal pre-performance, and determine the abnormal performance corresponding to the abnormal pre-performance as the early warning reference performance.

[0012] In some embodiments disclosed by the present invention, the present invention also discloses an Internet of Things data monitoring system based on artificial intelligence model analysis, including: The first module is used to determine the positions of data acquisition nodes corresponding to Internet of Things data, analyze the relative positions of different data acquisition nodes, associate the data acquisition nodes belonging to the same functional space with each other, configure the way of collecting mapping points in a preset virtual functional space, and connect the virtual functional spaces according to the relative position relationship to obtain an Internet of Things data space performance model; The second module is used to determine the performance during an anomaly and the performance before an anomaly of the Internet of Things data space performance model based on historical device anomaly records, and record the combination of the two as the performance group during an anomaly; The third module is used to perform anomaly chain performance clustering and parameter change equivalent clustering on the performance during an anomaly in sequence, and classify the performance group during an anomaly based on the final clustering result to obtain several performance sets during an anomaly; The fourth module is used to analyze the characteristics of the performance before an anomaly in the performance group during an anomaly belonging to the same performance set during an anomaly, construct a feature recognition model, and use the feature recognition model to perform prior anomaly monitoring on real-time Internet of Things data.

[0013] The present invention discloses an Internet of Things data monitoring method and system based on artificial intelligence model analysis, which relates to the technical field of Internet of Things data analysis. It includes determining the positions of data acquisition nodes, analyzing relative relationships, associating nodes in the same functional space, mapping them to a virtual functional space and connecting them to construct an Internet of Things data space performance model; extracting the performance during an anomaly and the performance before an anomaly of the model based on historical anomaly records, and combining them into a performance group during an anomaly; in step S300, performing chain and parameter change clustering on the performance during an anomaly and classifying it into several performance sets during an anomaly; analyzing the characteristics of the performance before an anomaly, constructing a feature recognition model, and monitoring anomalies in real time. The above technical solution of the present invention improves the accuracy and real-time performance of anomaly early warning by identifying the performance before an anomaly in advance, and effectively reduces the device risk.

[0014] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Description of the Drawings

[0015] Figure 1 It is a method step diagram of the Internet of Things data monitoring method based on artificial intelligence model analysis disclosed in the embodiments of the present invention. Detailed Embodiments

[0016] The technical solution of the present invention will be further described below with reference to the accompanying drawings and embodiments.

[0017] The technical solution of the present invention will be clearly and completely described below in conjunction with the accompanying drawings and specific embodiments. It should be understood that the preferred embodiments described herein are only used to illustrate and explain the present invention, and should not be construed as limiting the protection scope of the present invention. Those skilled in the art can make some non-essential improvements and adjustments according to the content of the present invention below. In the present invention, unless otherwise clearly specified and limited, the technical terms used in the present invention should have the ordinary meaning understood by those skilled in the art of the present invention.

[0018] Embodiment:

[0019] The present invention discloses an Internet of Things data monitoring method based on artificial intelligence model analysis, including: Step S100, determining the positions of the data acquisition nodes corresponding to the Internet of Things data, analyzing the relative positions of different data acquisition nodes, associating the data acquisition nodes belonging to the same functional space with each other, configuring the way of collecting mapping points in a preset virtual functional space, and connecting the virtual functional spaces according to the relative position relationship to obtain an Internet of Things data space representation model.

[0020] The core objective of step S100 is to determine the positions of the Internet of Things data acquisition nodes, analyze their relative relationships, associate the nodes belonging to the same functional space, map them to a preset virtual functional space, and finally connect them through the relative position relationship to form an Internet of Things data space representation model. The principle of this process is to use spatial geometry and functional relevance to abstract the scattered data acquisition nodes in the physical world into a structured virtual space representation form, laying a foundation for subsequent anomaly analysis. Specifically, in step S101, first calculate the average coordinates of all collection mapping points, define them as the center point of the virtual functional space, and then construct mapping connection lines starting from the center point according to the distance direction of each mapping point relative to the center point, and set a parameter status display module at the other end of the connection line.

[0021] It aims to capture the internal connections between nodes through geometric means. For example, nodes that are closer in distance and have the same direction may have stronger functional correlations, and this correlation is quantitatively reflected through the mapped connection lines. Step S102 further introduces dynamics. According to the parameter changes (such as temperature, pressure, etc.) at the data acquisition points, it drives the change of the parameter state performance module, making the model not only a static spatial framework but also capable of reflecting the dynamic state of the data in real time. Combining the technical solutions throughout the text, the construction of this model is not only a space for anomaly monitoring, but also provides an operable data carrier for the subsequent extraction of anomaly manifestations (S200) and feature analysis (S400) through the setting of the parameter state performance module. Compared with traditional isolated data monitoring methods, S100 overcomes the defect of the lack of context correlation between data points through spatial modeling, enabling anomaly analysis to start from a global perspective and discover potential patterns and laws.

[0022] Step S200: Based on the historical device anomaly records, determine the anomaly manifestations and pre-anomaly manifestations of the Internet of Things data space performance model, and record the combination of the two as the anomaly manifestation group.

[0023] Step S200 determines the anomaly manifestations and pre-anomaly manifestations of the Internet of Things data space performance model based on historical device anomaly records, and combines the two into the anomaly manifestation. Its principle is to identify the typical state when the device anomaly occurs and the omen characteristics before the anomaly through retrospective analysis of historical data, so as to provide a data basis for the mining and early warning of anomaly patterns. Specifically, step S201 uses the historical anomaly parameter sequence and its corresponding historical normal parameter sequence to drive the parameter state performance module constructed in step S100 to generate anomaly manifestations (such as parameter fluctuation characteristics during anomalies) and pre-anomaly manifestations (such as progressive deviation from the normal state before anomalies). This process relies on time series analysis and causal reasoning, aiming to extract the dynamic evolution law of anomalies from historical data. For example, a sudden increase in the temperature of a device may be accompanied by a gradual change in the pressure parameter, and this correlation is captured through the change of the parameter state performance module. Combining the full text, the significance of this step is to provide raw materials for the anomaly clustering in step S300 and at the same time provide labeled data for the training of the feature recognition model in step S400. Compared with traditional methods, S200 not only focuses on the anomaly itself, but also realizes the analysis perspective from the result back to the cause through the extraction of pre-anomaly manifestations. Thus, it improves the real-time performance and accuracy of the system. In addition, through the combination of the anomaly manifestation group, S200 also provides the possibility for modeling the diversity and complexity of anomalies. For example, anomalies may show different development paths due to different precursor manifestations.

[0024] Step S300: Perform anomaly chain manifestation clustering and parameter change equivalent clustering on the manifestations during anomalies in sequence, and classify the manifestation groups during anomalies based on the final clustering results to obtain several sets of manifestations during anomalies.

[0025] In step S300, by performing anomaly chain clustering and parameter change equivalent clustering on the manifestations during anomalies in sequence, and classifying the manifestation groups during anomalies into several sets of manifestations during anomalies based on the clustering results. Its principle lies in using multi-dimensional clustering analysis to reveal the propagation laws and similarities of anomalies in the spatial and temporal dimensions, thereby organizing complex anomaly patterns into recognizable categories. Specifically, in step S301, first analyze the manifestations during anomalies of each parameter status manifestation module, calculate the anomaly chain parameters (step S3011) based on the relative distance between modules and the time difference of anomaly occurrence, so as to determine the chain relationship of anomalies (such as the order and time difference of anomaly propagation from one node to another). This process quantifies the propagation characteristics of anomalies through the mathematical expression of anomaly chain parameters (combining distance and time weights) to ensure that the clustering results have physical significance. In step S3012, further compare the groups of chain relationships. If the anomaly chain orders are the same, they are classified into the same category to complete the anomaly chain clustering. Then, in step S302, based on the initial results of the anomaly chain clustering, judge the similarity of the manifestations during anomalies between different modules through the parameter change equivalent parameters (steps S3021 - S302). For example, if the anomaly parameter change trends of multiple modules are the same, it is determined as parameter change equivalent, and finally a reference clustering set is formed. The clustering method of S300 makes full use of the spatial model of S100 and the anomaly manifestation data of S200. Through double clustering of chain and equivalence, it overcomes the limitation that anomaly analysis in traditional monitoring is too scattered. This method can not only identify anomalies of a single node, but also capture the collaborative anomaly patterns between multiple nodes, providing a classification basis for feature recognition in step S400. In addition, the clustering results also reveal the regularity of anomalies. For example, some anomalies may always propagate between specific nodes in a specific order, providing scientific support for the precise prevention and control of equipment risks.

[0026] Step S400: Analyze the pre-anomaly manifestations in the manifestation groups during anomalies belonging to the same set of manifestations during anomalies, construct a feature recognition model, and use the feature recognition model to perform prior anomaly monitoring on real-time Internet of Things data.

[0027] Step S400 constructs a feature recognition model by analyzing the features of the pre - anomaly manifestations concentrated during the same anomaly, and uses this model to monitor anomalies in real - time Internet of Things data. Its principle lies in using machine learning and pattern recognition technologies to extract key features (such as parameter change trends, fluctuation amplitudes, etc.) from the pre - anomaly manifestations, and realizing early warning of anomalies through the matching of real - time data and features. Specifically, step S401 maps the real - time Internet of Things data to the Internet of Things data space manifestation model of S100, generates real - time manifestations and extracts feature sequences, ensuring that the extracted features can fully represent the dynamic changes of the data. Step S402 then compares the real - time feature sequence with the pre - anomaly manifestation feature sequence extracted in S200. If the matching degree is high, the corresponding anomaly - time manifestation is used as a warning reference. S400 integrates the results of the foregoing steps: the space model of S100 provides a data framework, the pre - anomaly manifestations of S200 provide training data, and the clustering results of S300 provide guidance for the classification and screening of features. Compared with traditional lagging monitoring methods, S400 shifts the focus of monitoring from after the anomaly occurs to before the anomaly occurs by identifying pre - anomaly manifestations in advance, significantly improving the real - time performance and accuracy of early warning. For example, if the real - time data shows a progressive equivalence similar to the historical pre - anomaly features, the system can issue an alarm in advance, thus effectively reducing the risk of equipment failure. In addition, the flexibility of this method is also reflected in its scalability. The feature recognition model can be continuously optimized with the input of new data to ensure adaptability for long - term applications.

[0028] In some embodiments disclosed in the present invention, the method for constructing an Internet of Things data space manifestation model includes: Step S101, calculate the average coordinate of the mapping points of the acquisition mapping points in the virtual function space, identify the average coordinate of the mapping points as the function space center point of the virtual function space, determine the inter - point distance between each acquisition mapping point and the function space center point, and construct a mapping connection line starting from the function space center point based on the inter - point distance and relative direction between the acquisition mapping points and the function space center point, and set a parameter status manifestation module at the other end of the mapping connection line.

[0029] Step S102, determine the change of the corresponding parameter status manifestation module based on the parameter change of the data acquisition point.

[0030] In some embodiments disclosed in the present invention, the method for determining the change of the parameter status manifestation model includes: Step S1021, construct manifestation spheres of several diameters, and each manifestation sphere of a diameter corresponds to a preset parameter interval.

[0031] Among them, the method for determining the parameter interval includes: Step S1022: Perform normal change clustering, abnormal change clustering, and fault change clustering on the historical data change records of each data collection point to obtain a normal data change group, an abnormal data change group, and a fault data change group respectively.

[0032] Step S1023: Calculate the average of the normal data changes in the normal data change group to obtain the average normal data change representing the normal data changes, calculate the average of the abnormal data changes in the abnormal data change group to obtain the average abnormal data change representing the abnormal data changes, calculate the average of the fault data change group to obtain the average fault data change, align the average abnormal data change, the average normal data change, and the average fault data change, and based on a preset reference time series, determine the average abnormal parameters, average normal parameters, and average fault parameters at different time nodes, and form an average abnormal parameter sequence, an average normal parameter sequence, and an average fault parameter sequence.

[0033] Step S1024: Divide several initial parameter intervals for each parameter sequence, determine the parameter quantity density mapped by the parameter sequence to each initial parameter interval. If the parameter quantity density is greater than or equal to a preset value, mark the initial parameter interval as a parameter interval to be concerned. If there are greater than or equal to a preset number of parameter intervals to be concerned in several consecutive initial parameter intervals, fuse the heads and tails of the several consecutive initial parameter intervals to obtain a new parameter interval.

[0034] In some embodiments disclosed by the present invention, the method for determining the abnormal time performance and the pre-abnormal performance of the Internet of Things data space performance model includes: Step S201: Based on the historical device abnormal records, determine several historical abnormal parameter sequences, determine the historical normal parameter sequences before each historical abnormal parameter sequence, and drive the corresponding parameter state performance module based on the corresponding historical normal parameter sequence and historical abnormal parameter sequence to obtain the abnormal time performance and the pre-abnormal performance of the Internet of Things data space performance model.

[0035] In some embodiments disclosed by the present invention, the method for sequentially performing abnormal chain clustering and parameter change equivalence clustering on the abnormal time performance in the abnormal time performance group includes: Step S301: Determine the abnormal time performance of each parameter state performance module when an abnormality occurs, and determine the abnormal chain performance of the parameter state performance module based on the relative distance between the parameter state performance modules and the occurrence time difference of the abnormal time performance. The chain performance includes the parameter state performance modules that chain to show abnormal time performance, the sequence of abnormal time performance, and the occurrence time difference.

[0036] Step S302: Based on the abnormal chain manifestations of different parameter state manifestation modules in the Internet of Things data space manifestation model, perform abnormal chain clustering on the manifestations during abnormal times of the Internet of Things data space manifestation model to obtain an initial clustering set. Compare the manifestations during abnormal times of each Internet of Things data space manifestation model in the initial clustering set, including comparing the manifestations during abnormal times of each parameter state manifestation module, and determine the equivalent parameters of the manifestations during abnormal times between each pair of relative parameter state manifestation modules. If the equivalent parameters of the manifestations during abnormal times are greater than or equal to the preset value, it is determined that the parameter changes between the corresponding parameter state manifestation modules are equivalent. If all pairs of relative parameter state manifestation modules are determined to have equivalent parameter changes, then classify the manifestations during abnormal times of the Internet of Things data space manifestation model again to obtain a reference clustering set.

[0037] In some embodiments disclosed by the present invention, the method for performing abnormal chain clustering on the manifestations during abnormal times of the Internet of Things data space manifestation model further includes: Step S3011: Based on the relative distance between parameter state manifestation modules and the difference amount of occurrence times of manifestations during abnormal times, determine the abnormal chain parameter between the manifestations during abnormal times of parameter state manifestation modules. Based on the abnormal chain parameter, determine whether there is an abnormal chain relationship between parameter state manifestation modules, and record the combination of parameter state manifestation modules with an abnormal chain relationship as a chain relationship group.

[0038] Step S3012: Compare the chain relationship groups between Internet of Things data space manifestation models. If the abnormal chain order of parameter state manifestation modules between chain relationship groups is equivalent, it is determined that the corresponding chain relationship groups are equivalent in manifestation. If all chain relationship groups are equivalent in manifestation, then classify the manifestations during abnormal times of the Internet of Things data space manifestation model into one category.

[0039] Among them, the expression for calculating the abnormal chain parameter between the manifestations during abnormal times of parameter state manifestation modules is: .

[0040] Among them, is the abnormal chain parameter, is the relative distance influence adjustment coefficient, is the preset maximum relative distance, is the relative distance between parameter state manifestation modules, is the occurrence time difference amount influence adjustment coefficient, is the preset maximum occurrence time difference amount, is the occurrence time difference amount of the manifestations during abnormal times of parameter state manifestation modules, is the abnormal chain parameter adjustment constant.

[0041] In some embodiments disclosed by the present invention, the method for determining equivalent parameters in abnormal performance between each relative parameter status performance module includes: Step S3021: Align the abnormal performance of the parameter status performance module, and determine the performance difference amount at different time nodes according to the preset time series relationship. If the performance difference amount is less than or equal to the preset value, it is determined that the corresponding time node is an equivalent performance time node.

[0042] Step S3022: Determine the equivalent parameters in abnormal performance based on the proportion of equivalent time nodes occupied by the equivalent performance time nodes in all time nodes.

[0043] In some embodiments disclosed by the present invention, the method for performing prior abnormal monitoring on real-time Internet of Things data by using a feature recognition model includes: Step S401: Map the real-time Internet of Things data to the Internet of Things data space performance model in real time, obtain the real-time performance of the Internet of Things data space performance model, and perform feature extraction on the real-time performance to obtain a real-time performance feature sequence. Step S402: Compare the real-time performance feature sequence with the abnormal pre-performance feature sequences of different abnormal pre-performances, and based on the comparison result, determine the matching abnormal pre-performance, and determine the abnormal performance corresponding to the abnormal pre-performance as the early warning reference performance.

[0044] In some embodiments disclosed by the present invention, the present invention also discloses an Internet of Things data monitoring system based on artificial intelligence model analysis, including: The first module is used to determine the positions of the data acquisition nodes corresponding to the Internet of Things data, analyze the relative positions of different data acquisition nodes, associate the data acquisition nodes belonging to the same functional space with each other, configure the acquisition mapping points in the preset virtual functional space, and connect the virtual functional spaces according to the relative position relationship to obtain the Internet of Things data space performance model.

[0045] The second module is used to determine the abnormal performance and abnormal pre-performance of the Internet of Things data space performance model based on the historical device abnormal records, and record the combination of the two as the abnormal performance group.

[0046] The third module is used to perform abnormal chain performance clustering and parameter change equivalent clustering on the abnormal performance in sequence, and classify the abnormal performance group based on the final clustering result to obtain several abnormal performance sets.

[0047] The fourth module is used to perform feature analysis on the abnormal pre-performances in the abnormal performance groups belonging to the same abnormal performance set, construct a feature recognition model, and use the feature recognition model to perform prior abnormal monitoring on the real-time Internet of Things data.

[0048] The present invention discloses an Internet of Things data monitoring method and system based on artificial intelligence model analysis, which relates to the technical field of Internet of Things data analysis. It includes determining the positions of data acquisition nodes, analyzing the relative relationships, associating nodes in the same functional space, mapping them to a virtual functional space and connecting them to construct an Internet of Things data space representation model; according to historical anomaly records, extracting the representations during and before anomalies of the model, and combining them into a set of representations during anomalies; in step S300, performing clustering on the representations during anomalies for chain and parameter changes, and classifying them into several sets of representations during anomalies; analyzing the feature characteristics before anomalies, constructing a feature recognition model, and monitoring anomalies in real time. The above technical solution of the present invention improves the accuracy and real-time performance of anomaly early warning by identifying the representations before anomalies in advance, and effectively reduces the equipment risks.

[0049] Through the description of the above embodiments, those skilled in the art can clearly understand that the present invention can be implemented through hardware, or can be implemented by means of software plus a necessary general hardware platform. Based on such an understanding, the technical solution of the present invention can be embodied in the form of a software product, which can be stored in a non-volatile storage medium (which can be a CD-ROM, a USB flash drive, a mobile hard disk, etc.), and includes several instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in various implementation scenarios of the present invention.

[0050] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit them. Although the present invention has been described in detail with reference to the preferred embodiments, those of ordinary skill in the art should understand that they can still modify or equivalently replace the technical solutions of the present invention, and these modifications or equivalent replacements cannot make the modified technical solutions deviate from the spirit and scope of the technical solutions of the present invention.

Claims

1. An Internet of Things data monitoring method based on artificial intelligence model analysis, characterized in that, Including: Step S100: Determine the positions of the data acquisition nodes corresponding to the Internet of Things data, analyze the relative positions of different data acquisition nodes, associate the data acquisition nodes belonging to the same functional space with each other, configure the method of collecting mapping points in a preset virtual functional space, and connect the virtual functional spaces according to the relative position relationship to obtain an Internet of Things data space performance model; Step S200: Based on historical device anomaly records, determine the performance during anomaly and the performance before anomaly of the Internet of Things data space performance model, and record the combination of the two as the anomaly performance group; Step S300: Perform anomaly chain performance clustering and parameter change equivalent clustering on the performance during anomaly in sequence, and classify the anomaly performance group based on the final clustering result to obtain several anomaly performance sets; Step S400: Analyze the performance before anomaly in the anomaly performance group belonging to the same anomaly performance set, construct a feature recognition model, and use the feature recognition model to perform prior anomaly monitoring on real-time Internet of Things data.

2. The method for monitoring Internet of Things data based on artificial intelligence model analysis according to claim 1, wherein The method for constructing an Internet of Things data space performance model includes: Step S101: Calculate the average coordinate of the mapping points of the collection mapping points in the virtual functional space, recognize the average coordinate of the mapping points as the center point of the functional space of the virtual functional space, determine the distance between each collection mapping point and the center point of the functional space, and construct a mapping connection line starting from the center point of the functional space based on the distance between the collection mapping point and the center point of the functional space and the relative direction, and set a parameter status performance module at the other end of the mapping connection line; Step S102: Determine the change of the corresponding parameter status performance module based on the parameter change of the data acquisition point.

3. The method for monitoring Internet of Things data based on artificial intelligence model analysis according to claim 2, wherein The method for determining the change of the parameter status performance model includes: Step S1021: Construct performance spheres of several diameters, and each performance sphere of a diameter corresponds to a preset parameter interval; Among them, the method for determining the parameter interval includes: Step S1022: Perform normal change clustering, abnormal change clustering, and fault change clustering on the historical data change records of each data acquisition point to obtain a normal data change group, an abnormal data change group, and a fault data change group respectively; Step S1023: Perform average calculation on the normal data changes in the normal data change group to obtain the average normal data change representing the normal data changes, perform average calculation on the abnormal data changes in the abnormal data change group to obtain the average abnormal data change representing the abnormal data changes, perform average calculation on the fault data change group to obtain the average fault data change, align the average abnormal data change, the average normal data change, and the average fault data change, and determine the average abnormal parameters, average normal parameters, and average fault parameters at different time nodes based on a preset reference time series, and form an average abnormal parameter sequence, an average normal parameter sequence, and an average fault parameter sequence; Step S1024: Divide several initial parameter intervals for each parameter sequence, determine the parameter quantity density mapped by the parameter sequence for each initial parameter interval. If the parameter quantity density is greater than or equal to the preset value, mark the initial parameter interval as a parameter interval that needs attention. If there are greater than or equal to the preset number of parameter intervals that need attention among several consecutive initial parameter intervals, fuse the heads and tails of the several consecutive initial parameter intervals to obtain a new parameter interval.

4. The method for monitoring Internet of Things data based on artificial intelligence model analysis according to claim 2, wherein, The method for determining the abnormal-time performance and pre-abnormal performance of the Internet of Things data space representation model includes: Step S201: Based on historical device abnormal records, determine several historical abnormal parameter sequences, determine the historical normal parameter sequences before each historical abnormal parameter sequence, and drive the corresponding parameter state performance module based on the corresponding historical normal parameter sequence and historical abnormal parameter sequence to obtain the abnormal-time performance and pre-abnormal performance of the Internet of Things data space representation model.

5. The method for monitoring Internet of Things data based on artificial intelligence model analysis according to claim 2, wherein The method for sequentially performing abnormal chain clustering and parameter change equivalence clustering on the abnormal-time performance in the abnormal-time performance group includes: Step S301: Determine the abnormal-time performance of each parameter state performance module when an abnormality occurs, and determine the abnormal chain performance of the parameter state performance module based on the relative distance between the parameter state performance modules and the time difference amount of the occurrence of the abnormal-time performance. The chain performance includes the parameter state performance modules that show abnormalities in a chain, the sequence of abnormal-time performance, and the time difference amount of occurrence; Step S302: Based on the abnormal chain performance of different parameter state performance modules in the Internet of Things data space representation model, perform abnormal chain clustering on the abnormal-time performance of the Internet of Things data space representation model to obtain an initial clustering set. Compare the abnormal-time performance of each Internet of Things data space representation model in the initial clustering set with each other, including comparing the abnormal-time performance of each parameter state performance module, and determine the equivalent parameters of the abnormal-time performance between each pair of parameter state performance modules. If the equivalent parameters of the abnormal-time performance are greater than or equal to the preset value, it is determined that the parameter changes between the corresponding parameter state performance modules are equivalent. If all pairs of parameter state performance modules are determined to have equivalent parameter changes, re-classify the abnormal-time performance of the Internet of Things data space representation model to obtain a reference clustering set.

6. The method for monitoring Internet of Things data based on artificial intelligence model analysis according to claim 5, characterized in that The method for performing abnormal chain clustering on the abnormal-time performance of the Internet of Things data space representation model also includes: Step S3011: Based on the relative distance between the parameter state performance modules and the time difference amount of the occurrence of the abnormal-time performance, determine the abnormal chain parameters between the abnormal-time performances of the parameter state performance modules. Based on the abnormal chain parameters, determine whether there is an abnormal chain relationship between the parameter state performance modules, and record the combination of the parameter state performance modules with an abnormal chain relationship as a chain relationship group; Step S3012: Compare the chain relationship groups between the IoT data space representation models. If the abnormal chain orders of the parameter status representation modules are the same between the chain relationship groups, it is determined that the corresponding chain relationship groups are equivalent to each other. If all the chain relationship groups are equivalent, classify the abnormal representations of the IoT data space representation models into one category; Among them, the expression for calculating the abnormal chain parameters between the abnormal representations of the parameter status representation modules is: ； Among them, is the abnormal chain parameter, is the relative distance influence adjustment coefficient, is the preset maximum relative distance, is the relative distance between parameter status display modules, is the occurrence time difference influence adjustment coefficient, is the preset maximum occurrence time difference, is the occurrence time difference of the abnormal performance of the parameter status display module, is the abnormal chain parameter adjustment constant.

7. The method for monitoring Internet of Things data based on artificial intelligence model analysis according to claim 5, characterized in that, The method for determining the equivalent parameters of the abnormal representations between each pair of parameter status representation modules includes: Step S3021: Align the abnormal representations of the parameter status representation modules, and determine the performance difference amounts at different time nodes according to the preset time series relationship. If the performance difference amount is less than or equal to the preset value, it is determined that the corresponding time node is an equivalent performance time node; Step S3022: Determine the equivalent parameter of the abnormal representation based on the proportion of the equivalent time nodes of the performance equivalent time nodes in all time nodes.

8. The method for monitoring Internet of Things data based on artificial intelligence model analysis according to claim 1, wherein The method for performing prior abnormal monitoring on real-time IoT data using a feature recognition model includes: Step S401: Map the real-time IoT data to the IoT data space representation model in real time to obtain the real-time representation of the IoT data space representation model, and perform feature extraction on the real-time representation to obtain a real-time representation feature sequence; Step S402: Compare the real-time representation feature sequence with the abnormal pre-representation feature sequences of different abnormal pre-representations, and based on the comparison results, determine the matching abnormal pre-representations, and regard the abnormal representations corresponding to the abnormal pre-representations as warning reference representations.

9. An Internet of Things data monitoring system based on artificial intelligence model analysis, characterized in that, Including: The first module is used to determine the positions of the data acquisition nodes corresponding to the IoT data, analyze the relative positions of different data acquisition nodes, associate the data acquisition nodes belonging to the same functional space with each other, configure the way of collecting mapping points in the preset virtual functional space, and connect the virtual functional spaces according to the relative position relationship to obtain the IoT data space representation model; The second module is used to determine the abnormal representation and the abnormal pre-representation of the IoT data space representation model based on the historical device abnormal records, and record the combination of the two as the abnormal representation group; The third module is used to perform abnormal chain performance clustering and parameter change equivalence clustering on the abnormal representations in sequence, and classify the abnormal representation groups based on the final clustering results to obtain several abnormal representation sets; The fourth module is used to perform feature analysis on the abnormal pre-representations in the abnormal representation groups belonging to the same abnormal representation set, construct a feature recognition model, and use the feature recognition model to perform prior abnormal monitoring on real-time IoT data.

Images