Iot-based intelligent interconnection device operation data processing method and system

Abstract

The application provides a kind of based on the Internet of Things intelligent interconnection equipment operation data processing method and system, by constructing equipment association graph, for each equipment node in the equipment association graph, the operation data sequence of the equipment node in the preset period is collected, target equipment operation state vector is generated based on the operation data sequence, based on the equipment association graph and the target equipment operation state vector, abnormal propagation simulation is carried out, the potential propagation path and path influence range of abnormal state between equipment nodes are determined;According to the potential propagation path and the path influence range, a set of equipment operation intervention strategies is generated, the set of equipment operation intervention strategies is input into a strategy verification model, and a strategy execution effect evaluation result is output, and the priority intervention instruction in the set of equipment operation intervention strategies is adjusted according to the strategy execution effect evaluation result to improve the accuracy and reliability of intelligent interconnection equipment operation data processing as a whole.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of data processing, in particular to a smart interconnection device operation data processing method and system based on Internet of Things. BACKGROUND

[0002] With the rapid development of Internet of Things technology, smart interconnection devices are widely used in various fields. Effective processing of the operation data of these devices is crucial to ensure the stable operation of the system. For example, as a common smart interconnection device in homes and public facilities, the smart interconnection water faucet realizes water data collection, remote control, and abnormal alarm functions through the integration of sensors and communication modules. Effective processing of its operation data is crucial to ensure water safety and stable operation of the device. Currently, the state of the device is monitored by collecting its operating parameters, and whether there is an anomaly is mainly determined by a simple parameter threshold. When analyzing the correlation between devices, a device network is constructed mainly based on physical connection or simple communication relationship. This makes it difficult to fully reflect the overall situation and potential risks of device operation. Simple device correlation construction often fails to accurately reflect the complex functional dependency relationship between devices and the actual impact of data interaction, leading to deviations in identifying abnormal propagation paths and assessing the impact range, and thus the generated intervention strategies lack pertinence and priority considerations, affecting the operation reliability and processing efficiency of the overall system. SUMMARY

[0003] The present application provides a smart interconnection device operation data processing method and system based on Internet of Things.

[0004] In a first aspect, the present application provides a smart interconnection device operation data processing method based on Internet of Things, comprising:

[0005] Constructing a device correlation graph, the device correlation graph containing multiple device nodes and node connection relationships. Each device node corresponds to a smart interconnection device in the Internet of Things, and the node connection relationship represents the data interaction frequency and functional dependency between devices.

[0006] For each device node in the device correlation graph, collect the operation data sequence of the device node within a preset time period, generate a target device operation state vector based on the operation data sequence, and the target device operation state vector contains device operation parameter features and parameter fluctuation features.

[0007] Based on the device correlation graph and the target device operation state vector, simulate the propagation of anomalies, determine the potential propagation path and path impact range of the abnormal state between device nodes.

[0008] According to the potential propagation path and the path influence range, a device operation intervention strategy set is generated, and the device operation intervention strategy set contains priority intervention instructions for different propagation paths;

[0009] The device operation intervention strategy set is input into a strategy verification model, a strategy execution effect evaluation result is output through the strategy verification model, and the priority intervention instructions in the device operation intervention strategy set are adjusted according to the strategy execution effect evaluation result.

[0010] In a second aspect, an embodiment of the present application provides a computer system, comprising:

[0011] A memory, wherein the memory stores a computer program;

[0012] A processor, configured to load the computer program to implement the Internet of Things-based smart interconnected device operation data processing method as described above.

[0013] The Internet of Things-based smart interconnected device operation data processing method provided by the present application constructs a device association graph to contain device nodes and node connection relationships, wherein the node connection relationships represent the data interaction frequency and the functional dependence degree between devices at the same time. Through this two-dimensional association relationship description, the actual association state of the device in the Internet of Things system can be more accurately reflected. The running data sequence of each device node is collected and a target device operation state vector containing device operation parameter characteristics and parameter fluctuation characteristics is generated. Through the fusion of static parameter characteristics and dynamic change characteristics, the current stability and change trend of device operation can be comprehensively reflected. Based on the device association graph and the target device operation state vector, abnormal propagation simulation is performed. Combined with the structural association between devices and the state characteristics of each device, the potential propagation path and the path influence range of the abnormal state between device nodes can be more accurately determined. According to the potential propagation path and the path influence range, a device operation intervention strategy set containing priority intervention instructions for different propagation paths is generated. Through the priority setting of the path difference, it can be ensured that the intervention resources are focused on the high-risk propagation path, and the pertinence and effectiveness of the intervention strategy are improved. The device operation intervention strategy set is input into a strategy verification model, and the priority intervention instructions are adjusted through the strategy execution effect evaluation result output by the model. The dynamic iterative optimization of the intervention strategy can be realized, the adaptability of the intervention strategy to the actual abnormal propagation scene is enhanced, and thus the accuracy and reliability of the smart interconnected device operation data processing are improved as a whole. BRIEF DESCRIPTION OF DRAWINGS

[0014] In order to make the technical solutions in the embodiments of the present application or the prior art clearer, the accompanying drawings needed in the embodiments or prior art description will be briefly introduced. Obviously, the accompanying drawings in the following description only aim to some embodiments of the present application, and for those skilled in the art, other drawings can be obtained without creative effort.

[0015] Figure 1 is a flowchart of a smart interconnected device operation data processing method based on Internet of Things provided by an embodiment of the present application.

[0016] Figure 2 is a composition schematic diagram of a computer system provided by an embodiment of the present application. DETAILED DESCRIPTION

[0017] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are only some embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative effort fall within the protection scope of the present application.

[0018] Please refer to Figure 1 , Figure 1 is a flowchart of a smart interconnected device operation data processing method based on Internet of Things provided by an embodiment of the present application. The method can be executed by a computer system, and can include the following steps.

[0019] Step S100: constructing a device association graph, the device association graph containing a plurality of device nodes and node connection relationships, each device node corresponding to a smart interconnected device in Internet of Things, and the node connection relationship representing data interaction frequency and functional dependence degree between devices.

[0020] The device association graph is a graphical structure for describing the relationship between smart interconnected devices in Internet of Things. The device node is a basic element of the graph, each device node corresponding to a specific smart interconnected device in Internet of Things, such as a smart Internet of Things faucet. The device node carries the relevant information of the device, and the node connection relationship is an important part of the graph, reflecting the data interaction frequency and functional dependence degree between devices. The data interaction frequency reflects the frequency of data transmission between devices, and the functional dependence degree represents the dependence of the normal operation of one device on another device.

[0021] As an implementation manner, in step S100, the device association graph is constructed, which can specifically include the following steps S110-S170.

[0022] Step S110: Collecting the basic information of all smart interconnected devices in the Internet of Things, which includes device identification, device type and device installation location.

[0023] The basic information of the device is the basis for identifying, classifying and positioning the device. The device identification is a unique identifier for each smart interconnected device, which is used to accurately distinguish different devices in the system. In practical applications, the device identification can be the serial number, MAC address, etc. of the device. For example, the serial number of the smart interconnected faucet is its unique identity, through which the corresponding device can be accurately found in the system. The device type clearly defines the function and purpose of the device, for example, the smart interconnected faucet belongs to the water-using device type, and different types of devices have different roles and operating modes in the system. The device installation location records the specific location of the device in the physical space. For the smart interconnected faucet, its installation location may be the kitchen, bathroom, etc. Collecting these basic information can be achieved in various ways. For device identification and device type information, the device's communication interface can be used to read from the device's configuration file using network protocols such as HTTP, MQTT, etc. For example, by sending a specific request instruction to the smart interconnected faucet, its serial number and device type information can be obtained. For device installation location information, it can be manually entered into the system through on-site investigation, or it can be automatically obtained using GIS technology combined with the device's positioning module (such as GPS).

[0024] Step S120: Establishing an initial device node set based on device identification, wherein each device node contains device type and device installation location information of the corresponding device.

[0025] After obtaining the basic information of the device, an initial device node set is established based on the device identification. Each device node not only contains the device identification, but also associates the device type and device installation location information. In this way, in the device association graph, each node can accurately represent a specific smart interconnected device and its related attributes. Specifically, an empty set is created to store device nodes. Then, all collected device basic information is traversed, and for each piece of basic information, the device identification is taken as the key, and the device type and device installation location information are taken as the value, encapsulated into a device node object, and the object is added to the set. For example, for the smart interconnected faucet, its device identification is used as the index, and the device type (water-using device) and installation location (kitchen) information are associated with it to form a device node. In this way, the uniqueness of each device node in the set can be ensured, and the management and operation of the device node can be facilitated.

[0026] Step S130: Collect data interaction records between each device node in a preset historical period, the data interaction records including interaction initiating device identifier, interaction receiving device identifier and interaction timestamp.

[0027] The data interaction records reflect the data transmission between devices in a certain time, and the preset historical period is a time range set according to actual needs, for example, it can be set to the past week, a month, etc. The interaction initiating device identifier and the interaction receiving device identifier in the data interaction records clearly show the source and destination devices of data transmission, and the interaction timestamp records the specific time of data interaction. Collecting these data interaction records can be achieved through network monitoring devices or device self-log recording functions. For example, deploying a traffic monitoring device in the network to monitor the network communication between devices in real time, and recording the initiating device identifier, receiving device identifier and timestamp of each data interaction. For the smart internet water faucet, if there is data interaction with the water quality monitoring device, the traffic monitoring device can record the relevant information of each interaction. In addition, the device itself can also record the data interaction log, and by reading the log file of the device, the required data interaction records can be extracted.

[0028] Step S140: Statistic the data interaction times between each device node pair in unit time, and take the data interaction times as the data interaction frequency value between device nodes.

[0029] The data interaction times between device node pairs in unit time are counted in order to quantify the data interaction frequency between devices. The unit time can be selected according to actual conditions, such as every minute, every hour, etc. Specifically, the interaction records between the set device node pairs are selected from the collected data interaction records. Then, according to the time range division, the data interaction times between the device node pairs in unit time are counted. For example, taking every hour as the unit time, the data interaction times between the smart internet water faucet and the water quality monitoring device in each hour are counted. The counted interaction times are taken as the data interaction frequency value between device nodes. The larger this value is, the more frequent the data interaction between devices is, and the closer the correlation is.

[0030] Step S150: Analyze the function description information of each device node, determine the function dependency relationship between device nodes, and generate a function dependency value based on the function dependency relationship.

[0031] The function description information of the device node usually contains the working principle of the device, input and output requirements, etc. Through the analysis of these information, the functional dependency relationship between the device nodes can be determined. For example, the water quality adjustment function of the intelligent water faucet depends on the water quality data provided by the water quality monitoring device, which indicates that there is a functional dependency relationship between the intelligent water faucet and the water quality monitoring device. After determining the functional dependency relationship, a functional dependency value needs to be generated to quantify the degree of dependency. The generation of the functional dependency value can be determined according to the functional logic of the device and the actual business requirements. For example, if the intelligent water faucet completely depends on the data of the water quality monitoring device to adjust the water quality, the functional dependency value between them can be set to a higher value; if it is only partially dependent, the functional dependency value is correspondingly reduced. The specific determination method can be through expert evaluation, rule matching, etc.

[0032] Step S160: Weighted combination of the data interaction frequency value and the functional dependency value to obtain the association strength value between the device nodes.

[0033] The association strength value comprehensively considers the data interaction frequency and the functional dependency degree between the device nodes, and more comprehensively reflects the association relationship between the devices. Weighted combination means that different weights are given to the data interaction frequency value and the functional dependency value, and then they are linearly combined. The determination of the weight needs to be adjusted according to the actual situation, for example, if more attention is paid to the data interaction frequency between the devices, a higher weight can be given to the data interaction frequency value; if more emphasis is placed on the functional dependency relationship, a higher weight is given to the functional dependency value. Specifically, first, the weight of the data interaction frequency value and the weight of the functional dependency value are determined. Then, the data interaction frequency value is multiplied by its corresponding weight, and the functional dependency value is multiplied by its corresponding weight. Finally, the two results are added to obtain the association strength value between the device nodes.

[0034] Step S170: When the association strength value is greater than the preset association threshold value, a connection relationship is established between the corresponding device nodes; all device nodes and node connection relationships are integrated to generate a device association map.

[0035] The preset association threshold is a preset critical value for determining whether there is a strong association relationship between device nodes. When the association strength value between device nodes is greater than the preset association threshold, it means that the association between them is relatively close, and a connection relationship needs to be established between the corresponding device nodes. Establishing a connection relationship can be achieved by adding an edge in the graph, with the two ends of the edge connecting two device nodes. Then, all device nodes and established connection relationships are integrated to form a complete device association graph. The integration process can be to store the device nodes and connection relationships in a graph database, or to display them in a graphical way. For example, using the graph database Neo4j to store the device association graph, taking the device nodes as the vertices in the graph and the connection relationships as the edges in the graph, through the query and operation functions of the graph database, the device association graph can be conveniently managed and analyzed.

[0036] Step S200: For each device node in the device association graph, collect the running data sequence of the device node within a preset time period, generate a target device running state vector based on the running data sequence, and the target device running state vector contains device running parameter features and parameter fluctuation features.

[0037] The running data sequence is a record set of the running state of the device within a preset time period, reflecting the working condition of the device within a period of time. The preset time period can be set according to actual needs, such as one day, one week, etc. The target device running state vector is a quantitative representation of the device running state, which can more comprehensively describe the running state of the device by extracting device running parameter features and parameter fluctuation features. The device running parameter features reflect the basic statistical characteristics of the device running parameters, and the parameter fluctuation features reflect the changes of the device running parameters over time. By generating the target device running state vector, a strong basis can be provided for subsequent anomaly detection and analysis.

[0038] As an implementation manner, step S200, for each device node in the device association graph, collecting the running data sequence of the device node within a preset time period, generating a target device running state vector based on the running data sequence, can specifically include steps S210-S240:

[0039] Step S210: For each device node in the device association graph, collect the running data sequence of the device node within a preset time period, and the running data sequence contains running parameter values corresponding to multiple timestamps.

[0040] The running data sequence is a specific embodiment of the running state of the device within a preset period of time, which is composed of running parameter values corresponding to multiple time stamps. Each time stamp corresponds to a time, and the running parameter value reflects the running state of the device at that time. For example, for a smart internet faucet, the running parameters can include water flow speed, water temperature, etc. Collecting these running data can be achieved through the device's own sensors and data acquisition module. The device's sensors monitor the changes in running parameters in real time, and the monitored data is recorded through the data acquisition module while adding the corresponding time stamp. After the end of the preset period, these recorded data are sorted in chronological order into a running data sequence. For example, within a day, the sensors of the smart internet faucet collect water flow speed and water temperature data every certain period of time (e.g., one minute) and record the corresponding time stamp, eventually forming a running data sequence containing multiple time stamps and running parameter values.

[0041] Step S220: feature extraction is performed on the running data sequence to obtain device running parameter features, which include mean features, variance features, skewness features and kurtosis features of each running parameter.

[0042] Feature extraction is a process of extracting key features from the running data sequence that can reflect the running state of the device. The mean feature in the device running parameter feature reflects the average level of the running parameter, the variance feature reflects the dispersion degree of the running parameter, the skewness feature describes the asymmetry of the running parameter distribution, and the kurtosis feature represents the peak degree of the running parameter distribution. By extracting these features, the running characteristics of the device can be better understood.

[0043] As an implementation, step S220, feature extraction is performed on the running data sequence to obtain device running parameter features, which can specifically include steps S221-S224:

[0044] Step S221: extract parameter sub-sequences corresponding to each running parameter from the running data sequence, each parameter sub-sequence containing all parameter values of the running parameter within the preset period of time.

[0045] The running data sequence contains the values of multiple running parameters. In order to extract features for each running parameter respectively, the parameter values corresponding to each running parameter need to be extracted to form a parameter sub-sequence. For example, in the running data sequence of a smart internet faucet, all parameter values of water flow speed are extracted to form a water flow speed parameter sub-sequence, and all parameter values of water temperature are extracted to form a water temperature parameter sub-sequence. The extraction process can be achieved by traversing the running data sequence, screening out the corresponding values according to the identification of the running parameter, and arranging them in chronological order.

[0046] Step S222: Perform outlier detection processing on the parameter subsequence, replace the abnormal data points whose parameter values exceed the preset N times standard deviation range with the median of the parameter subsequence, N ≥ 1.

[0047] The outlier detection processing is to remove abnormal noise in the running data and improve the accuracy of feature extraction. The preset N times standard deviation range is a threshold range set according to actual conditions, which is used to determine whether a data point is an outlier. When the parameter value exceeds this range, the data point is considered abnormal. In this step, the abnormal data points are replaced with the median of the parameter subsequence. The median is the middle value in the parameter subsequence, which has good robustness to outliers. Specifically, first calculate the mean and standard deviation of the parameter subsequence. Then, according to the preset N value, determine the judgment range of outliers. Next, traverse the parameter subsequence and mark the data points that exceed the range as outliers. Finally, replace these outliers with the median of the parameter subsequence. For example, for the water flow speed parameter subsequence of the intelligent Internet of Things faucet, calculate its mean and standard deviation, and assume that N is set to 3, replace the data points that exceed the mean plus or minus 3 times the standard deviation range with the median of the subsequence.

[0048] Step S223: Calculate the arithmetic mean of the processed parameter subsequence, and take the arithmetic mean as the mean feature of the running parameter; calculate the standard deviation of the processed parameter subsequence, and take the standard deviation as the variance feature of the running parameter; calculate the first ratio of the third central moment of the processed parameter subsequence to the cube of the standard deviation, and take the first ratio as the skewness feature of the running parameter; calculate the second ratio of the fourth central moment of the processed parameter subsequence to the fourth power of the standard deviation, and take the second ratio as the kurtosis feature of the running parameter.

[0049] The parameter subsequence after outlier processing can more accurately reflect the running state of the device, and by calculating these statistical features, the characteristics of the device running parameters can be further analyzed. The arithmetic mean is the sum of all values in the parameter subsequence divided by the number of values, which reflects the average level of the running parameter. The standard deviation measures the dispersion of data in the parameter subsequence, and the larger the standard deviation, the more dispersed the data. The third central moment and the fourth central moment are statistical quantities used to describe the shape of the data distribution. The skewness feature reflects the asymmetry of the data distribution, and positive skewness indicates that the data distribution is right-skewed, and negative skewness indicates that the data distribution is left-skewed. The kurtosis feature indicates the degree of peak of the data distribution, and the larger the kurtosis value, the more concentrated the data distribution around the mean. For example, for the water flow speed parameter subsequence of the intelligent Internet of Things faucet, calculate the arithmetic mean as the mean feature, the standard deviation as the variance feature, the ratio of the third central moment to the cube of the standard deviation as the skewness feature, and the ratio of the fourth central moment to the fourth power of the standard deviation as the kurtosis feature.

[0050] Step S224: Normalizing the mean feature, variance feature, skewness feature and kurtosis feature, mapping each feature value to a preset numerical interval, and arranging the normalized mean feature, variance feature, skewness feature and kurtosis feature of all running parameters according to a preset parameter order to obtain the device running parameter feature.

[0051] The normalization processing is to unify the range of different feature values to a preset numerical interval, which is convenient for subsequent processing and comparison. The preset numerical interval can be set according to actual needs, for example, [0, 1]. The normalization processing can adopt various methods, such as minimum-maximum normalization, Z-score normalization, etc. Taking the minimum-maximum normalization as an example, for each feature value, first find the minimum and maximum values of the feature in all running parameters, then subtract the minimum value from the feature value, and divide by the difference between the maximum value and the minimum value to obtain the normalized feature value. The normalized mean feature, variance feature, skewness feature and kurtosis feature of all running parameters are arranged according to the preset parameter order to form a vector, which is the device running parameter feature. For example, for the flow rate and water temperature of the intelligent Internet of Things faucet, the normalized mean feature, variance feature, skewness feature and kurtosis feature of the two running parameters are arranged in the order of flow rate first and water temperature last to obtain the device running parameter feature.

[0052] Step S230: Difference calculation is performed on the running parameter values of adjacent time stamps in the running data sequence to obtain a parameter fluctuation sequence, and a parameter fluctuation feature is generated based on the parameter fluctuation sequence. The parameter fluctuation feature includes fluctuation amplitude feature, fluctuation frequency feature and fluctuation trend feature.

[0053] The parameter fluctuation sequence reflects the change of the device running parameter with time. By difference calculation on the running parameter values of adjacent time stamps in the running data sequence, the parameter fluctuation sequence can be obtained. The parameter fluctuation feature generated based on the parameter fluctuation sequence can more comprehensively describe the fluctuation characteristics of the device running parameter. The fluctuation amplitude feature represents the size of the running parameter fluctuation, the fluctuation frequency feature reflects the frequency of the running parameter fluctuation, and the fluctuation trend feature reflects the direction and duration of the running parameter fluctuation.

[0054] As an implementation manner, in step S230, difference calculation is performed on the running parameter values of adjacent time stamps in the running data sequence to obtain a parameter fluctuation sequence, and a parameter fluctuation feature is generated based on the parameter fluctuation sequence. Specifically, the following steps S231-S236 can be included:

[0055] Step S231: For each running parameter in the running data sequence, a parameter sub-sequence corresponding to the running parameter is extracted.

[0056] Similar to step S221, in order to analyze the fluctuation of each running parameter respectively, the parameter value corresponding to each running parameter needs to be extracted from the running data sequence to form a parameter sub-sequence. For example, for the running data sequence of the intelligent Internet of Things faucet, the parameter sub-sequences of the water flow speed and the water temperature are extracted respectively.

[0057] Step S232: The parameter values of adjacent time stamps in the parameter sub-sequence are calculated by backward difference, and the parameter fluctuation sub-sequence of the running parameter is obtained. Each element in the parameter fluctuation sub-sequence is the difference between the current time stamp parameter value and the previous time stamp parameter value.

[0058] For each time stamp in the parameter sub-sequence, the parameter value of the current time stamp is subtracted from the parameter value of the previous time stamp, and the difference obtained is taken as an element in the parameter fluctuation sub-sequence. For example, for the water flow speed parameter sub-sequence of the intelligent Internet of Things faucet, the water flow speed difference of adjacent time stamps is calculated in turn to form the parameter fluctuation sub-sequence of the water flow speed.

[0059] Step S233: The average value of the absolute values of all non-zero elements in the parameter fluctuation sub-sequence is calculated, and the average value is taken as the fluctuation amplitude feature of the running parameter.

[0060] The fluctuation amplitude feature reflects the average size of the running parameter fluctuation. By calculating the average value of the absolute values of all non-zero elements in the parameter fluctuation sub-sequence, the fluctuation amplitude feature of the running parameter can be obtained. Non-zero elements indicate that the running parameter has changed, and calculating the average value of the absolute values of these changes can more accurately reflect the amplitude of the fluctuation. For example, for the water flow speed parameter fluctuation sub-sequence of the intelligent Internet of Things faucet, the average value of the absolute values of all non-zero elements is calculated as the fluctuation amplitude feature of the water flow speed.

[0061] Step S234: The third ratio of the number of non-zero elements in the parameter fluctuation sub-sequence to the length of the parameter sub-sequence is calculated, and the third ratio is taken as the fluctuation frequency feature of the running parameter.

[0062] The fluctuation frequency feature reflects the frequency of the running parameter fluctuation. By calculating the ratio of the number of non-zero elements in the parameter fluctuation sub-sequence to the length of the parameter sub-sequence, the fluctuation frequency feature of the running parameter can be obtained. The more the number of non-zero elements, the more frequent the running parameter fluctuation. For example, for the water temperature parameter fluctuation sub-sequence of the intelligent Internet of Things faucet, the number of non-zero elements is counted, and divided by the length of the water temperature parameter sub-sequence to obtain the fluctuation frequency feature of the water temperature.

[0063] Step S235: The maximum length of continuous positive difference values or continuous negative difference values in the parameter fluctuation sub-sequence is identified, and the ratio of the maximum length to a preset length threshold is taken as the fluctuation trend feature of the running parameter.

[0064] The fluctuation trend feature reflects the direction and duration of the fluctuation of the operation parameter. The continuous occurrence of positive difference values indicates that the operation parameter has an upward trend, and the continuous occurrence of negative difference values indicates that the operation parameter has a downward trend. The maximum length of the continuous occurrence of positive difference values or the continuous occurrence of negative difference values in the parameter fluctuation sub-sequence is identified, and the ratio of this maximum length to a preset length threshold is taken as the fluctuation trend feature. The preset length threshold can be set according to actual conditions, and is used to measure the significance of the fluctuation trend. For example, for the water flow velocity parameter fluctuation sub-sequence of the intelligent Internet-of-Things faucet, the maximum length of the continuous occurrence of positive difference values or the continuous occurrence of negative difference values is found, and the ratio of this maximum length to a preset length threshold is taken as the fluctuation trend feature of the water flow velocity.

[0065] Step S236: The fluctuation amplitude feature, the fluctuation frequency feature, and the fluctuation trend feature are standardized to make the characteristic values of each feature be in the same order of magnitude, and the standardized fluctuation amplitude features, the standardized fluctuation frequency features, and the standardized fluctuation trend features of all operation parameters are arranged in a preset parameter order to obtain a parameter fluctuation feature.

[0066] The standardization processing is to unify the orders of magnitude of the fluctuation amplitude feature, the fluctuation frequency feature, and the fluctuation trend feature, so as to facilitate subsequent processing and comparison. The standardization processing can adopt a Z-score standardization method. For each characteristic value, first, the mean and the standard deviation of the feature in all operation parameters are calculated, then the characteristic value is subtracted from the mean and divided by the standard deviation to obtain the standardized characteristic value. The standardized fluctuation amplitude features, the standardized fluctuation frequency features, and the standardized fluctuation trend features of all operation parameters are arranged in a preset parameter order to form a vector, and this vector is the parameter fluctuation feature. For example, for the water flow velocity and the water temperature of the intelligent Internet-of-Things faucet, the standardized fluctuation amplitude features, the standardized fluctuation frequency features, and the standardized fluctuation trend features of the two operation parameters are arranged in the order of water flow velocity first and water temperature second to obtain the parameter fluctuation feature.

[0067] Step S240: The device operation parameter feature and the parameter fluctuation feature are spliced to obtain an initial device operation state vector, and the initial device operation state vector is standardized to obtain a target device operation state vector.

[0068] The device running parameter feature and the parameter fluctuation feature are spliced, that is, two vectors are sequentially connected to form a longer vector, and the vector is an initial device running state vector. The splicing operation can be realized by simple vector merging. The initial device running state vector is normalized, and the same method as Z-score normalization can be used to normalize each element in the vector, so that the elements in the vector are in the same order of magnitude, facilitating subsequent analysis and processing. The vector obtained after normalization is the target device running state vector. For example, for a smart Internet of Things faucet, the device running parameter feature and the parameter fluctuation feature are spliced into an initial device running state vector, and then the vector is normalized to obtain a target device running state vector.

[0069] Step S300: Based on the device association graph and the target device running state vector, abnormal propagation simulation is performed to determine the potential propagation path and path impact range of the abnormal state between device nodes.

[0070] Abnormal propagation simulation is to predict the propagation of abnormal states between devices in order to take appropriate intervention measures in advance. The device association graph provides the association relationship between devices, and the target device running state vector reflects the current running state of the device. By combining these two pieces of information, the propagation process of the abnormal state between device nodes can be simulated to determine the potential propagation path and path impact range.

[0071] As an implementation manner, in step S300, based on the device association graph and the target device running state vector, abnormal propagation simulation is performed to determine the potential propagation path and path impact range of the abnormal state between device nodes, which can specifically include steps S310-S350:

[0072] Step S310: From the device association graph, filter out abnormal device nodes that meet the abnormal condition of the target device running state vector.

[0073] Abnormal device nodes refer to device nodes whose running state is abnormal, and filtering abnormal device nodes is the first step of abnormal propagation simulation. The abnormal condition is a judgment standard set according to actual conditions, which is used to judge whether the running state of the device node is abnormal.

[0074] As an implementation manner, in step S310, from the device association graph, filter out abnormal device nodes that meet the abnormal condition of the target device running state vector, which can specifically include steps S311-S316:

[0075] Step S311: Extract the device running parameter feature and the parameter fluctuation feature from the target device running state vector of each device node.

[0076] The target device running state vector contains device running parameter characteristics and parameter fluctuation characteristics. In order to determine whether the device is abnormal, the two characteristics need to be extracted and analyzed. The extraction process can be realized by vector index operation. According to the positions of the device running parameter characteristics and the parameter fluctuation characteristics in the target device running state vector, they are extracted respectively. For example, for the target device running state vector of the intelligent Internet of Things faucet, the device running parameter characteristics and the parameter fluctuation characteristics are extracted according to the pre-defined index positions.

[0077] Step S312: Compare the device running parameter characteristics with the preset normal parameter characteristic range, and count the number of abnormal parameter characteristics that exceed the preset parameter characteristic range.

[0078] The preset normal parameter characteristic range is a value interval set according to the normal operation of the device, which is used to determine whether the device running parameter is normal. Each characteristic value in the extracted device running parameter characteristics is compared with the corresponding normal parameter characteristic range. If a characteristic value exceeds the range, the characteristic is considered abnormal. The number of abnormal parameter characteristics that exceed the preset parameter characteristic range is counted, which can reflect the degree of abnormality of the device running parameter. For example, the mean characteristic and the variance characteristic in the device running parameter characteristics of the intelligent Internet of Things faucet are compared with the corresponding normal parameter characteristic range respectively, and the number of characteristics exceeding the range is counted.

[0079] Step S313: Compare the parameter fluctuation characteristics with the preset normal fluctuation characteristic range, and count the number of abnormal fluctuation characteristics that exceed the normal fluctuation characteristic range.

[0080] Similar to the device running parameter characteristics, the preset normal fluctuation characteristic range is a value interval used to determine whether the parameter fluctuation characteristics are normal. Each characteristic value in the extracted parameter fluctuation characteristics is compared with the corresponding normal fluctuation characteristic range, and the number of abnormal fluctuation characteristics that exceed the range is counted. For example, the fluctuation amplitude characteristic and the fluctuation frequency characteristic in the parameter fluctuation characteristics of the intelligent Internet of Things faucet are compared with the corresponding normal fluctuation characteristic range respectively, and the number of characteristics exceeding the range is counted.

[0081] Step S314: Calculate the ratio of the number of abnormal parameter characteristics to the total number of device running parameter characteristics to obtain the parameter abnormality rate, and calculate the ratio of the number of abnormal fluctuation characteristics to the total number of parameter fluctuation characteristics to obtain the fluctuation abnormality rate.

[0082] The parameter anomaly rate reflects the degree of anomaly in the equipment's operating parameters, while the fluctuation anomaly rate reflects the degree of anomaly in the fluctuation of the equipment's operating parameters. By calculating these two ratios, the anomaly status of the equipment can be assessed more quantitatively. For example, for a smart IoT faucet, the parameter anomaly rate is obtained by dividing the number of abnormal parameter features by the total number of equipment operating parameter features; the fluctuation anomaly rate is obtained by dividing the number of abnormal fluctuation features by the total number of parameter fluctuation features.

[0083] Step S315: Multiply the parameter anomaly rate by the first weighting coefficient to obtain the parameter anomaly score; multiply the fluctuation anomaly rate by the second weighting coefficient to obtain the fluctuation anomaly score; add the parameter anomaly score and the fluctuation anomaly score to obtain the overall equipment anomaly score.

[0084] The first and second weighting coefficients are weight values ​​set according to the actual situation, used to adjust the importance of parameter anomaly rate and fluctuation anomaly rate in the overall equipment anomaly score. Multiplying the parameter anomaly rate by the first weighting coefficient yields the parameter anomaly score; multiplying the fluctuation anomaly rate by the second weighting coefficient yields the fluctuation anomaly score. These two scores are then added together to obtain the overall equipment anomaly score. For example, for a smart IoT faucet, assuming the first weighting coefficient is 0.6 and the second weighting coefficient is 0.4, multiplying the parameter anomaly rate by 0.6 gives the parameter anomaly score, multiplying the fluctuation anomaly rate by 0.4 gives the fluctuation anomaly score, and then adding the two together gives the overall equipment anomaly score.

[0085] Step S316: When the overall score of device anomalies is greater than the preset anomaly threshold, the corresponding device node is identified as an abnormal device node.

[0086] The preset anomaly threshold is a pre-defined critical value used to determine whether a device is malfunctioning. When the overall anomaly score of a device exceeds the preset threshold, it indicates that the device's operating status is abnormal, and the corresponding device node is identified as an abnormal device node. For example, for a smart IoT faucet, if its overall anomaly score exceeds the preset threshold, the device node corresponding to that smart IoT faucet is identified as an abnormal device node.

[0087] Step S320: Taking the abnormal device node as the propagation starting point, construct an initial propagation path set based on the connection relationship between nodes in the device association graph. The initial propagation path set contains all possible paths starting from the abnormal device node.

[0088] The initial propagation path set is a set of all paths that the abnormal state can propagate, taking the abnormal device node as the propagation starting point. According to the connection relationship between nodes in the device association graph, all possible paths starting from the abnormal device node can be found. The specific implementation process can adopt a graph traversal algorithm, such as depth-first search (DFS), breadth-first search (BFS), etc. Taking depth-first search as an example, starting from the abnormal device node, recursive search is performed along the connection relationship between nodes, and all possible paths are recorded. These paths are added to the initial propagation path set. For example, for the smart internet water faucet, according to the connection relationship between it and other device nodes in the device association graph, all possible paths starting from it are found, and the initial propagation path set is constructed.

[0089] Step S330: For each path in the initial propagation path set, the association influence coefficient of each device node on the path is calculated, and the association influence coefficient is determined based on the connection relationship between adjacent nodes in the path and the target device running state vector.

[0090] The association influence coefficient reflects the degree of association between device nodes on the path and the degree of influence on abnormal propagation. Its calculation comprehensively considers the connection relationship between adjacent nodes in the path and the target device running state vector.

[0091] As an implementation manner, step S330, for each path in the initial propagation path set, the association influence coefficient of each device node on the path is calculated, which can specifically include the following steps S331-S336:

[0092] Step S331: For each path in the initial propagation path set, determine the device node sequence on the path, and the device node sequence contains all device nodes starting from the propagation starting point.

[0093] For each path in the initial propagation path set, arrange the device nodes on the path in order from the propagation starting point to the end point to form a device node sequence. This sequence clearly shows the order of device nodes on the path, providing a basis for subsequent calculation of the association influence coefficient. For example, for a path starting from the smart internet water faucet, the device nodes on the path are arranged in sequence to form a device node sequence.

[0094] Step S332: For each device node in the device node sequence except the propagation starting point, extract the connection relationship data of the device node and the previous device node from the device association graph, and the connection relationship data includes the data interaction frequency value and the function dependency value.

[0095] The connection relationship data reflects the degree of association between adjacent device nodes. In order to calculate the association influence coefficient, these data need to be extracted. For each device node in the device node sequence except the propagation starting point, find the connection relationship data of the device node and the previous device node from the device association graph, including the data interaction frequency value and the function dependency value. For example, for a certain device node in the device node sequence, extract the data interaction frequency value and the function dependency value of the device node and the previous device node from the device association graph.

[0096] Step S333: Normalize the data interaction frequency value to a preset numerical interval to obtain a standardized interaction frequency, and map the function dependency value to a corresponding weight coefficient. The weight coefficient is positively correlated with the function dependency value.

[0097] The normalization processing is to unify the range of the data interaction frequency value to a preset numerical interval, so as to facilitate subsequent calculation and comparison. The preset numerical interval can be set according to actual needs, for example, [0, 1]. The normalization processing can adopt the minimum-maximum normalization method. The function dependency value is mapped to a corresponding weight coefficient, and the weight coefficient is positively correlated with the function dependency value, that is, the greater the function dependency value, the greater the corresponding weight coefficient. For example, for the data interaction frequency value, the minimum-maximum normalization method is used to normalize it to the [0, 1] interval to obtain the standardized interaction frequency; for the function dependency value, a mapping function is used to convert it to a corresponding weight coefficient.

[0098] Step S334: Perform a product operation on the standardized interaction frequency and the weight coefficient to obtain the path dependence coefficient of the device node.

[0099] The path dependence coefficient reflects the degree of association between the device node and the previous device node. By multiplying the standardized interaction frequency and the weight coefficient, the path dependence coefficient of the device node can be obtained. For example, for a certain device node in the device node sequence, multiply the standardized interaction frequency of the device node and the previous device node with the corresponding weight coefficient to obtain the path dependence coefficient of the device node.

[0100] Step S335: Extract the parameter fluctuation feature from the target device running state vector of the device node, calculate the vector norm value of the parameter fluctuation feature, and the vector norm value is obtained by square sum operation of each fluctuation feature component.

[0101] The vector modulus of the parameter fluctuation feature reflects the fluctuation degree of the operating parameters of the device node. The parameter fluctuation feature is extracted from the target device operating state vector of the device node, and then the modulus of the vector is calculated. The calculation method of the vector modulus is to add the square of each component in the vector, and then take the square root of the sum. For example, for a device node in the sequence of device nodes, the parameter fluctuation feature is extracted from the target device operating state vector thereof, the square of the fluctuation amplitude feature, the fluctuation frequency feature, and the like are added, and then the vector modulus is obtained by taking the square root.

[0102] Step S336: The path dependence coefficient is multiplied by the vector modulus to obtain the correlation influence coefficient of the device node. The correlation influence coefficients of all device nodes in the sequence of device nodes are accumulated to obtain the total correlation influence coefficient of the path.

[0103] The correlation influence coefficient comprehensively considers the correlation degree between device nodes and the fluctuation of the operating parameters of the device nodes. The correlation influence coefficient of the device node is obtained by multiplying the path dependence coefficient and the vector modulus. The correlation influence coefficients of all device nodes in the sequence of device nodes are accumulated to obtain the total correlation influence coefficient of the path. The total correlation influence coefficient reflects the possibility and influence degree of the abnormal propagation on the path. For example, for a path starting from the intelligent Internet of Things faucet, the correlation influence coefficients of each device node on the path are calculated, and then they are added to obtain the total correlation influence coefficient of the path.

[0104] Step S340: The initial propagation path set is screened according to the correlation influence coefficient, and the paths with the correlation influence coefficient greater than a preset threshold value are retained to obtain the potential propagation path.

[0105] The preset threshold value is a pre-set critical value for screening paths with a greater possibility of abnormal propagation. The initial propagation path set is screened according to the correlation influence coefficient, and the paths with the correlation influence coefficient greater than the preset threshold value are retained, which are the potential propagation paths. For example, for all paths in the initial propagation path set, the paths with the correlation influence coefficient greater than the preset threshold value are screened to form the potential propagation path set.

[0106] Step S350: For each potential propagation path, the influence range index of the potential propagation path is calculated based on the target device operating state vector of each device node on the potential propagation path, and the device node set corresponding to the influence range index is determined as the path influence range.

[0107] The influence range index reflects the influence degree of the potential propagation path on the device nodes, and the path influence range can be determined by calculating the influence range index.

[0108] As an implementation, step S350, for each potential propagation path, based on the target device running state vectors of each device node on the potential propagation path, calculate the influence range index of the potential propagation path, which can specifically include the following steps S351-S357:

[0109] Step S351: For each potential propagation path, obtain the target device running state vectors of all device nodes on the potential propagation path.

[0110] In order to calculate the influence range index of the potential propagation path, it is necessary to obtain the target device running state vectors of all device nodes on the path. These vectors contain the running state information of the device and are the basis for calculating the influence range index. The corresponding vector can be obtained from the database or data structure storing the target device running state vector through the identification of the device node. For example, for a potential propagation path, the target device running state vector of each device node is obtained from the database according to the identification of the device node on the path.

[0111] Step S352: Extract parameter fluctuation features from each target device running state vector, including fluctuation amplitude features and fluctuation frequency features.

[0112] The parameter fluctuation features reflect the fluctuation of the device running parameters, and the parameter fluctuation features are extracted from each target device running state vector, including the fluctuation amplitude features and the fluctuation frequency features. The extraction process can be realized by vector index operation, and the fluctuation amplitude features and the fluctuation frequency features are extracted according to their positions in the target device running state vector. For example, for the target device running state vector of the smart internet water faucet, the fluctuation amplitude features and the fluctuation frequency features are extracted according to the pre-defined index positions.

[0113] Step S353: Calculate the product of the absolute value of the fluctuation amplitude feature and the fluctuation frequency feature in each parameter fluctuation feature to obtain the fluctuation influence value of the device node; and arithmetically average the fluctuation influence values of all device nodes to obtain the path average fluctuation value.

[0114] The fluctuation influence value reflects the comprehensive influence degree of the device node operation parameter fluctuation. The fluctuation influence value of each device node can be obtained by calculating the product of the absolute value of the fluctuation amplitude feature and the fluctuation frequency feature in each parameter fluctuation feature. The path average fluctuation value is obtained by arithmetically averaging the fluctuation influence values of all device nodes. The path average fluctuation value reflects the average level of the device node operation parameter fluctuation on the potential propagation path. For example, for each device node on the potential propagation path, the product of the absolute value of the fluctuation amplitude feature and the fluctuation frequency feature is calculated to obtain the fluctuation influence value. Then the fluctuation influence values of all device nodes are added and divided by the number of device nodes to obtain the path average fluctuation value.

[0115] Step S354: The number of device nodes on the potential propagation path is counted, and the path average fluctuation value is multiplied by the number of device nodes to obtain the path influence basis value.

[0116] The path influence basis value considers the number of device nodes on the path and the average level of the device node operation parameter fluctuation. The path influence basis value can be obtained by multiplying the path average fluctuation value by the number of device nodes. For example, for a potential propagation path, the number of device nodes on the path is counted, and the path average fluctuation value is multiplied by the number to obtain the path influence basis value.

[0117] Step S355: The path length of the path is obtained from the device association graph, and the path length is the number of connection edges between device nodes on the path. The reciprocal of the path length is calculated, and the path influence basis value is multiplied by the reciprocal to obtain the preliminary influence range index.

[0118] The path length reflects the complexity of the path. The longer the path length, the more difficult the abnormal propagation may be. The path length of the path is obtained from the device association graph, which is the number of connection edges between device nodes on the path. The reciprocal of the path length is calculated, and the path influence basis value is multiplied by the reciprocal to obtain the preliminary influence range index. For example, for a potential propagation path, the path length of the path is obtained from the device association graph, the reciprocal of the length is calculated, and the path influence basis value is multiplied by the reciprocal to obtain the preliminary influence range index.

[0119] Step S356: Key device nodes are extracted from the path influence range of the potential propagation path.

[0120] A key device node is a device node in the device association graph that has more than a preset number of connection edges. The abnormality of a key device node can affect more devices, and the key device nodes are extracted from the path influence range of the potential propagation path. The criterion for determining a key device node is that the number of connection edges in the device association graph is more than a preset number threshold. For example, for a potential propagation path, the device nodes with more than a preset number of connection edges in the device association graph are filtered from the path influence range, and these device nodes are determined as key device nodes.

[0121] Step S357: The number of key device nodes is counted, and the preliminary influence range index is multiplied by the square root of the number of key device nodes to obtain the influence range index of the potential propagation path.

[0122] The number of key device nodes affects the influence range of the potential propagation path, and the preliminary influence range index is multiplied by the square root of the number of key device nodes to obtain the influence range index of the potential propagation path. This influence range index comprehensively considers the fluctuation of device nodes on the path, the path length, and the number of key device nodes. For example, for a potential propagation path, the number of key device nodes is counted, and the preliminary influence range index is multiplied by the square root of the number of key device nodes to obtain the influence range index of the potential propagation path.

[0123] Step S400: According to the potential propagation path and the path influence range, a device operation intervention strategy set is generated, and the device operation intervention strategy set includes priority intervention instructions for different propagation paths.

[0124] The device operation intervention strategy set is a series of intervention measures formulated to deal with the propagation of abnormal states. According to the potential propagation path and the path influence range, priority intervention instructions for different propagation paths can be generated. The priority intervention instructions clearly indicate the order and manner of intervention on the devices in different propagation paths.

[0125] As an implementation manner, in step S400, according to the potential propagation path and the path influence range, a device operation intervention strategy set is generated, which can specifically include the following steps S410-S480:

[0126] Step S410: For each potential propagation path, the device node set corresponding to the path influence range of the potential propagation path and the influence range index of the potential propagation path are obtained.

[0127] The device node set corresponding to the path influence range is a set of all device nodes that can be affected by the potential propagation path, and the influence range index reflects the influence degree of the potential propagation path. Through the identification of the potential propagation path, the corresponding device node set and influence range index can be obtained from the database or data structure storing the path influence range and influence range index. For example, for a potential propagation path, the device node set corresponding to the path influence range and the influence range index are obtained from the database according to the identification thereof.

[0128] Step S420: The influence range index is sorted in descending order to sort the potential propagation paths, and a path priority sequence is obtained.

[0129] The path priority sequence clearly defines the order of importance of the potential propagation paths, and the influence range index is sorted in descending order to sort the potential propagation paths. The larger the influence range index, the higher the priority of the path. Various sorting algorithms can be used for sorting, such as bubble sort, quicksort, etc. For example, for all potential propagation paths, their influence range indexes are compared and arranged in descending order to obtain a path priority sequence.

[0130] Step S430: For each potential propagation path in the path priority sequence, the intervention priority of each device node in the device node set of the potential propagation path is determined according to the target device running state vector of each device node.

[0131] The intervention priority clearly defines the order of intervention of the device nodes on a potential propagation path, and the intervention priority of each device node can be determined according to the target device running state vector thereof.

[0132] As an implementation, step S430, the intervention priority of each device node in the device node set of the potential propagation path is determined according to the target device running state vector of each device node, which can specifically include steps S431-S435:

[0133] Step S431: For each device node in the device node set of the potential propagation path, the target device running state vector of the device node is obtained.

[0134] The target device running state vector contains the running state information of the device and is an important basis for determining the intervention priority. Through the identification of the device node, the corresponding vector can be obtained from the database or data structure storing the target device running state vector. For example, for each device node in the device node set of the potential propagation path, the target device running state vector of the device node is obtained from the database according to the identification thereof.

[0135] Step S432: Extract the parameter fluctuation feature from the target device running state vector, calculate the vector norm value of the parameter fluctuation feature, multiply the vector norm value by the first preset weight to obtain the fluctuation influence value.

[0136] The fluctuation influence value reflects the influence degree of the device node running parameter fluctuation. The parameter fluctuation feature is extracted from the target device running state vector, the norm value of the vector is calculated, the vector norm value is multiplied by the first preset weight to obtain the fluctuation influence value. The calculation method of the vector norm value is to add the square of each component in the vector, and then take the square root of the sum. For example, for a certain device node in the device node set, the parameter fluctuation feature is extracted from the target device running state vector of the device node, the square of the fluctuation amplitude feature, the fluctuation frequency feature and other components is added, and then the square root is taken to obtain the vector norm value. Then, the vector norm value is multiplied by the first preset weight to obtain the fluctuation influence value.

[0137] Step S433: Obtain the connection node quantity of the device node from the device association graph, the connection node quantity is the total number of other device nodes having a connection relationship with the device node, and the connection node quantity is multiplied by the second preset weight to obtain the connection influence value.

[0138] The connection influence value reflects the importance of the device node in the device association graph. The connection node quantity of the device node is obtained from the device association graph, and the connection node quantity is multiplied by the second preset weight to obtain the connection influence value. The more the connection node quantity, the more extensive the connection of the device node in the device association graph, and the abnormality of the device node may affect more devices. For example, for a certain device node in the device node set, the connection node quantity of the device node is obtained from the device association graph, and the connection node quantity is multiplied by the second preset weight to obtain the connection influence value.

[0139] Step S434: Obtain the position coefficient of the device node in the current path, the position coefficient is negatively correlated with the distance of the device node to the propagation starting point.

[0140] The position coefficient reflects the position importance of the device node in the potential propagation path, and the position coefficient is negatively correlated with the distance of the device node to the propagation starting point. That is, the closer the device node to the propagation starting point, the greater the position coefficient. The shortest path length of the device node to the propagation starting point can be calculated, and then a preset mapping function is used to convert the shortest path length into the position coefficient. For example, for a certain device node in the device node set, the shortest path length of the device node to the propagation starting point is calculated, and then a mapping function is used to convert the length into the position coefficient.

[0141] Step S435: Sum the fluctuation influence value, the connection influence value, and the position coefficient to obtain an intervention priority score of the device node, and sort the device nodes in the device node set according to the intervention priority score from high to low to obtain the intervention priority of each device node.

[0142] The intervention priority score comprehensively considers the running parameter fluctuation of the device node, the connection in the device association graph, and the position in the potential propagation path. The fluctuation influence value, the connection influence value, and the position coefficient are summed to obtain the intervention priority score of the device node. The device nodes in the device node set are sorted according to the intervention priority score from high to low, and the higher the score, the higher the intervention priority of the device node. For example, for all device nodes in the device node set of the potential propagation path, the intervention priority scores of the device nodes are calculated, and then the device nodes are sorted according to the scores from high to low to obtain the intervention priority of each device node.

[0143] Step S440: Based on the intervention priority and the position sequence of the device node in the path, an initial intervention instruction for the path is generated, and the initial intervention instruction includes a device node identifier, an intervention operation type, and an execution time window.

[0144] The initial intervention instruction specifies the specific content of the intervention on the device node. Based on the intervention priority and the position sequence of the device node in the path, the initial intervention instruction for the path can be generated. The intervention operation type can be set according to the specific situation of the device and the abnormal type, such as restarting the device, adjusting the parameter, etc. The execution time window refers to the time range of the intervention operation. For example, for a potential propagation path, according to the intervention priority and the position sequence of the device node, an initial intervention instruction including a device node identifier, an intervention operation type, and an execution time window is generated for each device node.

[0145] Step S450: The initial intervention instructions of all potential propagation paths are summarized in the order of the path priority sequence to obtain an initial intervention strategy set.

[0146] The initial intervention strategy set is a summary of the initial intervention instructions of all potential propagation paths. The initial intervention instructions of all potential propagation paths are summarized in the order of the path priority sequence to form the initial intervention strategy set. For example, for all potential propagation paths, their initial intervention instructions are sequentially added to a set in the order of the path priority sequence to obtain the initial intervention strategy set.

[0147] Step S460: The intervention instructions in the initial intervention strategy set are subjected to conflict detection, and the conflict detection includes device node conflict detection and time window conflict detection.

[0148] Conflict detection is to avoid the conflict between intervention instructions, and ensure the effective execution of intervention measures. Device node conflict detection refers to detecting whether there are multiple intervention instructions for the same device node, and time window conflict detection refers to detecting whether the time windows of intervention instructions for the same device node overlap. For example, for intervention instructions in the initial intervention strategy set, check whether there are multiple instructions for the same device node, and whether the execution time windows of these instructions overlap.

[0149] Step S470: When multiple intervention instructions for the same device node are detected, the intervention instruction with the highest intervention priority is retained; when the time windows of intervention instructions for the same device node overlap, the time window of the intervention instruction executed later is adjusted to make the time windows non-overlapping.

[0150] When multiple intervention instructions for the same device node are detected, the intervention instruction with the highest intervention priority is retained to ensure that the intervention on the device node is the most effective. When the time windows of intervention instructions for the same device node overlap, the time window of the intervention instruction executed later is adjusted to make the time windows non-overlapping, avoiding mutual interference between intervention operations. For example, for a certain device node, if there are multiple intervention instructions, compare their intervention priorities, and retain the instruction with the highest priority; if there is a time window overlap, adjust the time window of the instruction executed later so that it does not overlap with the time window of the previous instruction.

[0151] Step S480: Integrate the intervention instructions after conflict detection to generate a device operation intervention strategy set.

[0152] Integrate the intervention instructions after conflict detection to form a complete device operation intervention strategy set. The integration process can be to add the intervention instructions retained after conflict detection to a new set. For example, for intervention instructions after conflict detection, add them to a new set in sequence to obtain a device operation intervention strategy set.

[0153] Step S500: Input the device operation intervention strategy set into the strategy verification model, output the strategy execution effect evaluation result through the strategy verification model, and adjust the priority intervention instruction in the device operation intervention strategy set according to the strategy execution effect evaluation result.

[0154] The strategy verification model is used to evaluate the execution effect of the device operation intervention strategy set. The device operation intervention strategy set is input into the strategy verification model, and the model outputs the strategy execution effect evaluation result. According to the evaluation result, the priority intervention instruction in the device operation intervention strategy set can be adjusted to improve the effectiveness of the intervention strategy.

[0155] As an implementation, in step S500, the device operation intervention strategy set is input into the strategy verification model, the strategy execution effect evaluation result is output through the strategy verification model, and the priority intervention instruction in the device operation intervention strategy set is adjusted according to the strategy execution effect evaluation result. Specifically, the following steps S510-S550 can be included:

[0156] In step S510, the device operation intervention strategy set, the device correlation graph, and the target device operation state vector of each device node are input into the strategy verification model. The strategy verification model is a deep learning model trained based on historical intervention data, and includes a path simulation layer and an effect evaluation layer.

[0157] The strategy verification model is a deep learning model trained based on historical intervention data, which can evaluate the execution effect of the device operation intervention strategy set. The device operation intervention strategy set, the device correlation graph, and the target device operation state vector of each device node are input into the strategy verification model. The path simulation layer is used to simulate the execution process of the intervention strategy, and the effect evaluation layer is used to evaluate the execution effect of the intervention strategy. For example, the device operation intervention strategy set, the device correlation graph, and the target device operation state vector of each device node are input data, which are input into the input layer of the strategy verification model. The data is first processed in the path simulation layer.

[0158] In step S520, the path simulation layer of the strategy verification model is used to simulate the execution of the device operation intervention strategy set, and the abnormality mitigation rate and the device operation recovery time corresponding to each intervention instruction are output.

[0159] The path simulation layer simulates the execution process of the intervention strategy, and simulates the execution effect of the intervention instruction according to the device correlation graph and the target device operation state vector of each device node. In the simulation process, the path simulation layer predicts the propagation of abnormal states and the recovery of devices according to the correlation between devices and the operation state of devices. Finally, the abnormality mitigation rate and the device operation recovery time corresponding to each intervention instruction are output. The abnormality mitigation rate reflects the degree of mitigation of the intervention instruction to the abnormal state, and the device operation recovery time reflects the time required for the device to recover from the abnormal state to the normal state. For example, the path simulation layer uses a graph neural network structure, and the input data flows between the layers of the graph neural network. The input layer of the graph neural network receives the device correlation graph and the target device operation state vector of each device node. The intermediate layer simulates the interaction between devices and the abnormal propagation process through node feature aggregation and message passing mechanism. For each intervention instruction, the abnormality mitigation rate and the device operation recovery time are calculated according to the simulation result, and are output from the output layer.

[0160] Step S530: The effect evaluation layer of the strategy verification model performs weighted calculation on the abnormality mitigation rate and the device operation recovery time to generate an effect score of each intervention instruction, and performs arithmetic averaging on the effect scores of all intervention instructions to obtain a strategy execution effect score.

[0161] The effect evaluation layer performs weighted calculation on the abnormality mitigation rate and the device operation recovery time output by the path simulation layer to generate an effect score of each intervention instruction. Different weights are respectively assigned to the abnormality mitigation rate and the device operation recovery time, the abnormality mitigation rate is multiplied by the corresponding weight, the device operation recovery time is multiplied by the corresponding weight, and then the two results are added to obtain the effect score of each intervention instruction. The effect scores of all intervention instructions are arithmetically averaged to obtain a strategy execution effect score. For example, the effect evaluation layer adopts a fully connected neural network structure, the input layer receives the abnormality mitigation rate and the device operation recovery time output by the path simulation layer, the intermediate layer performs weighted summation and other calculations on the input data, and the output layer outputs the effect score of each intervention instruction. Finally, the effect scores of all intervention instructions are arithmetically averaged to obtain the strategy execution effect score from the output layer.

[0162] Step S540: If the strategy execution effect score is greater than the preset score threshold, the current device operation intervention strategy set is determined as the final intervention strategy; if the strategy execution effect score is less than or equal to the preset score threshold, the intervention instruction with the lowest effect score is selected from the device operation intervention strategy set, and the priority level of the intervention instruction is improved.

[0163] The preset score threshold is a pre-set critical value for judging whether the strategy execution effect meets the requirements. If the strategy execution effect score is greater than the preset score threshold, it indicates that the execution effect of the current device operation intervention strategy set is good, and the current device operation intervention strategy set is determined as the final intervention strategy. If the strategy execution effect score is less than or equal to the preset score threshold, it indicates that the strategy execution effect is poor, and the intervention instruction with the lowest effect score is selected from the device operation intervention strategy set, and the priority level of the intervention instruction is improved, so as to improve the overall execution effect. For example, the strategy execution effect score and the preset score threshold are compared. If the score is greater than the threshold, the current device operation intervention strategy set is taken as the final intervention strategy; if the score is less than or equal to the threshold, the intervention instruction with the lowest effect score is found, and the priority level of the intervention instruction is improved.

[0164] Step S550: The device operation intervention strategy set with the adjusted priority level is re-input into the strategy verification model, and the simulation and evaluation step is repeatedly executed until the strategy execution effect score is greater than the preset score threshold.

[0165] The device operation intervention strategy set with adjusted priority levels is input into the strategy verification model again for simulation evaluation. The simulation evaluation step is repeated until the strategy execution effect score is greater than the preset score threshold. For example, the device operation intervention strategy set with adjusted priority levels is taken as new input data and input into the strategy verification model again. The new strategy execution effect score is obtained after the path simulation layer and the effect evaluation layer are processed again. If the new score is still less than or equal to the preset score threshold, the priority level of the intervention instruction with the lowest effect score is adjusted again, and simulation evaluation is performed again until the score is greater than the preset score threshold.

[0166] As an implementation manner, the training process of the strategy verification model can include the following steps S501-S506:

[0167] Step S501: Obtain historical device operation intervention records, and the historical device operation intervention records include a historical device association graph, a historical device operation state vector, a historical intervention strategy set, and corresponding execution effect data.

[0168] The historical device operation intervention records are basic data for training the strategy verification model. By collecting the historical device operation intervention records, the historical device association graph, the historical device operation state vector, the historical intervention strategy set, and the corresponding execution effect data are obtained. These data can be obtained from log records of a device management system and data storage of a monitoring system. For example, historical device operation intervention records are extracted from a database of a device management system, including device association graphs, device operation state vectors, intervention strategy sets, and execution effect data of these intervention strategies, such as abnormality alleviation rates and device operation recovery times, in different time periods.

[0169] Step S502: Feature encoding is performed on each intervention instruction in the historical intervention strategy set to generate an intervention instruction feature vector; and the historical device operation state vector is standardized to obtain a standardized historical state vector.

[0170] Feature encoding is to convert the intervention instruction into a vector form that the model can process. Each intervention instruction in the historical intervention strategy set is feature encoded to generate an intervention instruction feature vector. Methods such as one-hot encoding and word embedding can be used to encode each attribute of the intervention instruction, such as intervention operation type, execution time window, etc. The historical device running state vector is standardized. Methods such as Z-score standardization are used to standardize each element in the vector so that the elements in the vector are in the same order of magnitude. For example, for each intervention instruction in the historical intervention strategy set, the intervention operation type is one-hot encoded, the execution time window is numerically processed, and the intervention instruction feature vector is combined. For the historical device running state vector, the mean and standard deviation are calculated, each element in the vector is subtracted from the mean, and then divided by the standard deviation to obtain the standardized historical state vector.

[0171] Step S503: Combine the historical device association graph, the standardized historical state vector, and the intervention instruction feature vector to construct a model training sample set, and divide the model training sample set into a training set, a validation set, and a test set.

[0172] The historical device association graph, the standardized historical state vector, and the intervention instruction feature vector are combined to form a complete training sample. All training samples are combined to construct a model training sample set. In order to evaluate the performance of the model and prevent overfitting, the model training sample set is divided into a training set, a validation set, and a test set. The division ratio can be set according to the actual situation, for example, the training set accounts for 70%, the validation set accounts for 15%, and the test set accounts for 15%. For example, the historical device association graph, the standardized historical state vector, and the intervention instruction feature vector are combined into a sample according to a certain format, and all samples are added to a list to form a model training sample set. Then, using a random division method, the sample set is divided into a training set, a validation set, and a test set.

[0173] Step S504: Construct a deep learning model containing a path simulation layer and an effect evaluation layer. The path simulation layer adopts a graph neural network structure, and the effect evaluation layer adopts a fully connected neural network structure.

[0174] The constructed strategy verification model includes a path simulation layer and an effect evaluation layer. The path simulation layer adopts a graph neural network structure, which can well process graph structure data such as a device association graph. The graph neural network includes an input layer, an intermediate layer, and an output layer. The input layer receives the device association graph and the state information of the device nodes. The intermediate layer simulates the interaction and abnormal propagation process between devices through node feature aggregation and message passing mechanism, and outputs the abnormal mitigation rate and the device operation recovery time corresponding to each intervention instruction. The effect evaluation layer adopts a fully connected neural network structure. The fully connected neural network can perform complex nonlinear transformation on the input data. The input layer of the fully connected neural network receives the abnormal mitigation rate and the device operation recovery time output by the path simulation layer. The intermediate layer includes multiple neuron layers, which process the input data through a series of weighted summation and activation function operations, and outputs the effect score of each intervention instruction.

[0175] For example, in the graph neural network of the path simulation layer, the input layer initializes the node features and edge features in the device association graph. The node features can include the type and running state of the device, and the edge features can represent the connection strength between devices. In the message passing process of the intermediate layer, each node updates its own features according to the information of its neighbor nodes. Specifically, the node collects the messages passed by the neighbor nodes, aggregates these messages (such as summation, averaging, etc.), and then updates its own features based on the current features. This process is iterated multiple times, so that the node can obtain more extensive graph structure information, thereby more accurately simulating the abnormal propagation situation. The output layer calculates and outputs the abnormal mitigation rate and the device operation recovery time based on the updated node features.

[0176] In the fully connected neural network of the effect evaluation layer, the input layer receives the two features of the abnormal mitigation rate and the device operation recovery time output by the path simulation layer. The neurons in the intermediate layer perform weighted summation on the input features, and then perform nonlinear transformation through an activation function (such as ReLU function). The ReLU function can set values less than 0 to 0 and keep values greater than 0 unchanged, which can introduce nonlinear factors and enhance the expression ability of the model. Multiple neuron layers will perform such operations in turn to process and abstract the input data layer by layer. Finally, the output layer calculates and outputs the effect score of each intervention instruction based on the output of the intermediate layer.

[0177] Step S505: training the constructed deep learning model using the training set, and optimizing the model parameters through a backpropagation algorithm.

[0178] The training process is to let the model learn the mapping relationship between the input data and the output data to improve the prediction accuracy of the model. The constructed deep learning model is trained using the training set, and the training process is an iterative optimization process. In each iteration, the samples in the training set are input into the model, the model calculates the output results according to the current parameters, and then compares the output results with the true labels of the samples to calculate the loss function value. The loss function is used to measure the difference between the model's output results and the true labels, and common loss functions include mean square error loss function, cross-entropy loss function, etc. Taking the mean square error loss function as an example, it calculates the average of the squares of the differences between the model output values and the true values.

[0179] Then, the gradient is calculated according to the loss function value using the backpropagation algorithm, which represents the rate of change of the loss function with respect to the model parameters. The backpropagation algorithm starts from the output layer and calculates the gradient layer by layer, passing the gradient information back to the parameters of each layer. According to the gradient information, the parameters of the model are updated using optimization algorithms such as stochastic gradient descent algorithm, Adam optimization algorithm, etc. The stochastic gradient descent algorithm updates the parameters according to the direction of the gradient and a preset learning rate, which controls the step size of each parameter update. Through continuous iteration of this process, the parameters of the model are gradually adjusted, so that the loss function value is continuously reduced, and the performance of the model is improved.

[0180] For example, in each iteration, one or a batch of samples are randomly selected from the training set and input into the model. Assuming that the selected samples contain historical device association graphs, standardized historical state vectors, and intervention instruction feature vectors, the model calculates the output abnormality mitigation rate and effect score according to these inputs. Compare the output abnormality mitigation rate with the true abnormality mitigation rate of the sample, and compare the output effect score with the true effect score to calculate the mean square error loss function value. Then, use the backpropagation algorithm to calculate the gradient of the loss function with respect to each parameter in the model (such as the weight matrix in the graph neural network, the bias term in the fully connected neural network, etc.). According to the gradient information, the Adam optimization algorithm is used to update these parameters, so that the model can more accurately output results in subsequent predictions.

[0181] Step S506: Monitor the overfitting situation in the model training process based on the validation set, stop training when the validation set loss no longer decreases; and perform performance evaluation on the trained deep learning model based on the test set, and when the evaluation index meets the preset requirements, determine the model as the strategy verification model.

[0182] Overfitting refers to the phenomenon that the model performs well on the training set but poorly on unseen data. To monitor overfitting during model training, a validation set is used to evaluate the model. After each training iteration, the samples in the validation set are input into the model, and the loss function value on the validation set is calculated. If the validation set loss no longer decreases or even starts to rise after multiple iterations, it indicates that the model may have overfitting, and training is stopped at this time.

[0183] After training is complete, the model is evaluated for performance using a test set. The test set is a separate data set from the training and validation sets, used to evaluate the model's generalization ability. Select appropriate evaluation metrics to measure the performance of the model, common evaluation metrics include accuracy, recall, F1 value, etc. Accuracy represents the proportion of correctly predicted samples to the total number of samples, recall represents the proportion of correctly predicted positive samples to the actual number of positive samples, and F1 value is the harmonic mean of accuracy and recall.

[0184] For example, during training, after completing a certain number of iterations (e.g. 10), the samples in the validation set are input into the model, and the mean square error loss function value of the validation set is calculated. If the validation set loss does not decrease after multiple iterations (e.g. 5), training is stopped. After training is complete, the samples in the test set are input into the model, and the difference between the model output anomaly mitigation rate and effect score and the true value is calculated, and the evaluation metrics are calculated based on these differences. If the evaluation metrics such as accuracy, recall and F1 value meet the pre-set requirements (e.g. accuracy reaches a certain threshold), the model is determined as a strategy verification model, which is used to evaluate the execution effect of the device running intervention strategy set in the future.

[0185] Through the above series of steps, the entire process of the smart interconnected device running data processing method based on the Internet of Things is completed. Starting from constructing the device association graph, the running data of the device is collected and feature extracted, the potential propagation path and path impact range are determined through anomaly propagation simulation, the device running intervention strategy set is generated, and finally the strategy verification model is used to evaluate and adjust the intervention strategy, forming a complete closed-loop system that can effectively process the running data of smart interconnected devices, timely detect device anomalies and take appropriate intervention measures to ensure stable operation of the device.

[0186] It can be understood that the various algorithms involved in the above introduction of the embodiments of the present application can be known from the related content in the prior art. In order to save space, the above introduction of the embodiments of the present application is not expanded too much. In addition, those skilled in the art can supplement the details according to the common knowledge in the art when implementing the scheme of the present application. For example, according to the common knowledge in the art, the normalization can be used to eliminate the dimensional conflict before feature fusion, the interpolation can be used to eliminate the dimensional difference, the threshold can be reasonably set according to the historical data, experience or business scene demand, the model can be trained based on the general model training method, the number of layers in the model structure can be set based on the actual needs, the activation function can be selected, and the like. The present application does not introduce the redundant implementation process in too much detail.

[0187] Please refer to Figure 2 , Figure 2 A structural schematic diagram of a computer system provided by the embodiments of the present application is shown in the figure. The computer system at least includes a processor 101, a communication interface 102 and a memory 103. The processor 101, the communication interface 102 and the memory 103 can be connected through a bus or other means. The processor 101 (or called Central Processing Unit, CPU) is the computing core and control core of the computer system, which can parse various instructions in the computer system and process various data of the computer system. The communication interface 102 can optionally include a standard wired interface, a wireless interface (such as WI-FI, a mobile communication interface, etc.), and can be used for transmitting and receiving data under the control of the processor 101; the communication interface 102 can also be used for the transmission and interaction of data in the computer system. The memory 103 is a memory device in the computer system, used for storing programs and data. It can be understood that the memory 103 here can include the built-in memory of the computer system, and of course can also include the extended memory supported by the computer system. The memory 103 provides a storage space which stores the operating system of the computer system, and the present application does not limit this.

[0188] In one embodiment, the processor 101 executes the computer program in the memory 103 to perform the data processing method of the intelligent interconnected device based on the Internet of Things provided by the embodiments of the present application.

Claims

1. A method for processing operational data of smart interconnected devices based on the Internet of Things, characterized in that, include: Construct a device association graph, which includes multiple device nodes and the connection relationships between nodes. Each device node corresponds to a smart interconnected device in the Internet of Things, and the connection relationships between nodes represent the frequency of data interaction and the degree of functional dependence between devices. For each device node in the device association map, the operating data sequence of the device node within a preset time period is collected, and a target device operating status vector is generated based on the operating data sequence. The target device operating status vector includes device operating parameter characteristics and parameter fluctuation characteristics. Based on the device association map and the target device operating state vector, anomaly propagation simulation is performed to determine the potential propagation path and the scope of influence of the path of the anomaly between device nodes. Based on the potential propagation path and the scope of influence of the path, a set of equipment operation intervention strategies is generated, which includes priority intervention instructions for different propagation paths; The set of equipment operation intervention strategies is input into the strategy verification model, and the strategy verification model outputs the strategy execution effect evaluation result. The priority intervention instructions in the set of equipment operation intervention strategies are adjusted according to the strategy execution effect evaluation result.

2. The method as described in claim 1, characterized in that, For each device node in the device association map, the step of collecting the operating data sequence of that device node within a preset time period and generating a target device operating state vector based on the operating data sequence includes: For each device node in the device association map, the operation data sequence of the device node within a preset time period is collected, and the operation data sequence contains operation parameter values ​​corresponding to multiple timestamps; Feature extraction is performed on the operational data sequence to obtain equipment operational parameter features, which include the mean, variance, skewness, and kurtosis features of each operational parameter. The difference between the values ​​of the running parameters at adjacent timestamps in the running data sequence is calculated to obtain the parameter fluctuation sequence. Based on the parameter fluctuation sequence, parameter fluctuation features are generated, which include fluctuation amplitude features, fluctuation frequency features, and fluctuation trend features. The device operating parameter features and the parameter fluctuation features are concatenated to obtain an initial device operating state vector, and the initial device operating state vector is standardized to obtain the target device operating state vector.

3. The method as described in claim 2, characterized in that, The step of extracting features from the operational data sequence to obtain equipment operational parameter features includes: Extract parameter subsequences corresponding to each operating parameter from the operating data sequence. Each parameter subsequence contains all parameter values ​​of the operating parameter within a preset time period. Outlier detection processing is performed on the parameter subsequence, and outlier data points whose parameter values ​​exceed a preset range of N times the standard deviation are replaced with the median of the parameter subsequence, where N≥1; The arithmetic mean of the processed parameter subsequence is calculated, and the arithmetic mean is used as the mean characteristic of the operating parameter; the standard deviation of the processed parameter subsequence is calculated, and the standard deviation is used as the variance characteristic of the operating parameter; the first ratio of the third central moment of the processed parameter subsequence to the cube of the standard deviation is calculated, and the first ratio is used as the skewness characteristic of the operating parameter; the second ratio of the fourth central moment of the processed parameter subsequence to the fourth power of the standard deviation is calculated, and the second ratio is used as the kurtosis characteristic of the operating parameter. The mean, variance, skewness, and kurtosis features are normalized to map each feature value to a preset numerical range. The normalized mean, variance, skewness, and kurtosis features of all operating parameters are then arranged in the order of preset parameters to obtain the equipment operating parameter features.

4. The method as described in claim 2, characterized in that, The step of calculating the difference between the operating parameter values ​​of adjacent timestamps in the operating data sequence to obtain a parameter fluctuation sequence, and generating parameter fluctuation features based on the parameter fluctuation sequence, includes: For each running parameter in the running data sequence, extract the parameter subsequence corresponding to that running parameter; The parameter fluctuation subsequence of the running parameter is obtained by performing backward difference calculation on the parameter values ​​of adjacent timestamps in the parameter subsequence, where each element in the parameter fluctuation subsequence is the difference between the current timestamp parameter value and the previous timestamp parameter value. Calculate the average of the absolute values ​​of all non-zero elements in the parameter fluctuation subsequence, and use the average value as the fluctuation amplitude characteristic of the operating parameter; The third ratio of the number of non-zero elements in the parameter fluctuation subsequence to the length of the parameter subsequence is calculated, and the third ratio is used as the fluctuation frequency characteristic of the operating parameter. Identify the maximum length of consecutive positive or negative differences in the parameter fluctuation subsequence, and use the ratio of the maximum length to a preset length threshold as the fluctuation trend feature of the operating parameter; The fluctuation amplitude characteristics, fluctuation frequency characteristics, and fluctuation trend characteristics are standardized to ensure that each characteristic value is on the same order of magnitude. The standardized fluctuation amplitude characteristics, fluctuation frequency characteristics, and fluctuation trend characteristics of all operating parameters are then arranged in a preset parameter order to obtain the parameter fluctuation characteristics.

5. The method as described in claim 1, characterized in that, The step of simulating anomaly propagation based on the device association map and the target device's operating state vector to determine the potential propagation paths and impact ranges of anomalies between device nodes includes: Filter out abnormal device nodes whose target device operating status vectors meet abnormal conditions from the device association map; Taking the abnormal device node as the propagation starting point, an initial propagation path set is constructed based on the connection relationship between nodes in the device association graph. The initial propagation path set contains all possible paths starting from the abnormal device node. For each path in the initial propagation path set, the association influence coefficient of each device node on the path is calculated. The association influence coefficient is determined based on the connection relationship between adjacent nodes in the path and the target device's operating state vector. The initial propagation path set is filtered based on the correlation influence coefficient, and paths with correlation influence coefficients greater than a preset threshold are retained to obtain potential propagation paths; For each potential propagation path, the influence range index of the propagation path is calculated based on the target device operating state vector of each device node on the propagation path, and the set of device nodes corresponding to the influence range index is determined as the path influence range.

6. The method as described in claim 5, characterized in that, For each path in the initial propagation path set, the correlation influence coefficient of each device node on that path is calculated, including: For each path in the initial propagation path set, a sequence of device nodes on that path is determined, the sequence of device nodes including all device nodes starting from the propagation origin; For each device node in the device node sequence except for the propagation starting point, the connection relationship data between the device node and the preceding device node is extracted from the device association graph. The connection relationship data includes data interaction frequency value and functional dependency value. The data interaction frequency value is normalized to a preset numerical range to obtain a standardized interaction frequency. The functional dependency value is mapped to a corresponding weight coefficient, and the weight coefficient is positively correlated with the functional dependency value. The path dependency coefficient of the device node is obtained by multiplying the standardized interaction frequency with the weight coefficient. Extract parameter fluctuation features from the target device operating state vector of the device node, calculate the vector magnitude of the parameter fluctuation features, and obtain the vector magnitude by taking the square root of the sum of squares of each fluctuation feature component; The path dependency coefficient is multiplied by the vector magnitude to obtain the association influence coefficient of the device node. The association influence coefficients of all device nodes in the device node sequence are summed to obtain the total association influence coefficient of the path.

7. The method as described in claim 5, characterized in that, For each potential propagation path, based on the target device operating state vector of each device node along that potential propagation path, the influence range index of that potential propagation path is calculated, including: For each potential propagation path, obtain the target device running state vector of all device nodes on that potential propagation path; Extract parameter fluctuation features from the operating state vector of each target device, wherein the parameter fluctuation features include fluctuation amplitude features and fluctuation frequency features; Calculate the product of the absolute value of the fluctuation amplitude feature and the fluctuation frequency feature in the fluctuation characteristics of each parameter to obtain the fluctuation impact value of that equipment node; take the arithmetic mean of the fluctuation impact values ​​of all equipment nodes to obtain the path average fluctuation value. The number of device nodes on the potential propagation path is counted, and the average fluctuation value of the path is multiplied by the number of device nodes to obtain the basic value of the path impact. The path length of the path is obtained from the device association graph, where the path length is the number of connecting edges between device nodes on the path; the reciprocal of the path length is calculated, and the basic value of the path influence is multiplied by the reciprocal to obtain the preliminary influence range index; Extract key device nodes from the path influence range of the potential propagation path. The key device nodes are device nodes whose number of connected edges in the device association graph exceeds a preset threshold. The number of key equipment nodes is counted, and the preliminary impact range index is multiplied by the square root of the number of key equipment nodes to obtain the impact range index of the potential propagation path.

8. The method as described in claim 1, characterized in that, The step of generating a set of device operation intervention strategies based on the potential propagation path and the scope of influence of the path includes: For each potential propagation path, obtain the set of device nodes corresponding to the path influence range of the potential propagation path and the influence range index of the potential propagation path; The potential propagation paths are sorted in descending order of the influence range index to obtain a path priority sequence; For each potential propagation path in the path priority sequence, the intervention priority of each device node is determined based on the target device operation state vector of each device node in the device node set of the potential propagation path. Based on the intervention priority and the position order of the device nodes in the path, an initial intervention instruction is generated for the path. The initial intervention instruction includes the device node identifier, the intervention operation type, and the execution time window. The initial intervention instructions for all potential propagation paths are summarized in order of path priority sequence to obtain the initial intervention strategy set; Conflict detection is performed on the intervention instructions in the initial intervention strategy set, and the conflict detection includes device node conflict detection and time window conflict detection. When multiple intervention commands targeting the same device node are detected, the intervention command with the highest priority is retained; when the time windows of intervention commands targeting the same device node overlap, the time window of the subsequently executed intervention command is adjusted to ensure that the time windows do not overlap. The intervention instructions after conflict detection are integrated to generate a set of equipment operation intervention strategies.

9. The method as described in claim 8, characterized in that, The step of determining the intervention priority of each device node based on the target device operating state vector of each device node in the device node set of the potential propagation path includes: For each device node in the set of device nodes in the potential propagation path, obtain the target device running state vector of that device node; Extract parameter fluctuation features from the target device's operating state vector, calculate the vector magnitude of the parameter fluctuation features, and multiply the vector magnitude by a first preset weight to obtain the fluctuation impact value; The number of connected nodes of the device node is obtained from the device association map. The number of connected nodes is the total number of other device nodes that are connected to the device node. The number of connected nodes is multiplied by the second preset weight to obtain the connection influence value. Obtain the position coefficient of the device node in the current path, where the position coefficient is negatively correlated with the distance from the device node to the propagation starting point; The fluctuation impact value, connection impact value, and position coefficient are summed to obtain the intervention priority score of the device node. The device nodes in the device node set are then sorted from high to low according to the intervention priority score to obtain the intervention priority of each device node.

10. A computer system, characterized in that, include: A memory, wherein a computer program is stored; A processor is used to load the computer program to implement the data processing method for the operation of smart interconnected devices based on the Internet of Things as described in any one of claims 1-9.

Images