Intelligent linkage management system based on IoT technology

Abstract

The application relates to the technical field of device control, in particular to an intelligent linkage management system based on IoT technology, which comprises a rule base construction unit, a control unit, a distribution deviation calculation unit and a rule base updating unit.The rule base construction unit is used for clustering historical environment parameters to obtain a plurality of clustering clusters and constructing a rule base according to the clustering clusters.The control unit is used for inputting real-time environment parameters at any moment into a fuzzy control algorithm to obtain control parameters.The distribution deviation calculation unit is used for calculating the distribution deviation of each clustering cluster.The rule base updating unit is used for calculating the environment stability of any clustering cluster when the distribution deviation of the clustering cluster is greater than a deviation threshold value, and updating the fuzzy rule of the clustering cluster in response to the environment stability being not greater than a stability threshold value.Through the technical scheme, the accuracy of device management in a smart park can be improved.

Description

Technical Field

[0001] This application relates to the field of equipment control technology, and more particularly to an intelligent linkage management system based on IoT technology. Background Technology

[0002] A smart park refers to a new type of park that achieves digital and intelligent management by comprehensively sensing, interconnecting in real time, and intelligently analyzing elements such as people, objects, equipment, and environment within the park, supported by next-generation information technologies such as IoT (Internet of Things), big data, and artificial intelligence.

[0003] Currently, patent application CN120370818A discloses an intelligent control method and system for electronic devices based on the Internet of Things (IoT). The method includes: collecting all data categories and control objectives of the target electronic device in the IoT, and analyzing the contribution of each data category to the control objective; setting Kalman filter parameters according to the contribution of the data categories to the control objective, and fusing all data categories to obtain fused features; defining fuzzy sets and membership functions based on the fused features to establish fuzzy control rules for the target electronic device; collecting all data of the target electronic device in the IoT in real time, performing fuzzy inference according to the fuzzy control rules, generating control commands, and intelligently controlling the target electronic device according to the control commands.

[0004] The above method defines fuzzy sets and membership functions based on the fusion features, and selects the type of membership function according to the characteristics of the fusion features, which can improve the pertinence of fuzzy reasoning. However, the elements such as people, things, equipment and environment in the smart park are constantly changing. The above method ignores the impact of these element changes on fuzzy reasoning, resulting in low accuracy of equipment management in the smart park. Summary of the Invention

[0005] To address the technical issue of low accuracy in the coordinated management of equipment within smart parks, this application provides an intelligent coordinated management system based on IoT technology, which can improve the accuracy of equipment management within smart parks.

[0006] In a first aspect, this application provides an intelligent linkage management system based on IoT technology. The management system includes: a rule base construction unit, used to cluster historical environmental parameters to obtain multiple clusters, and initialize each cluster as a fuzzy rule to construct a rule base; a control unit, used to input real-time environmental parameters at any time into a fuzzy control algorithm to obtain the activation degree of each fuzzy rule in the rule base, and obtain control parameters; a distribution deviation calculation unit, used to obtain the control error of the real-time environmental parameters, calculate the long-term centroid and short-term centroid of the cluster to which the real-time environmental parameters belong based on the control error, and use the Euclidean distance between the long-term centroid and the short-term centroid as the distribution deviation of the cluster; and a rule base update unit, used to calculate the environmental stability of the cluster when the distribution deviation of any cluster is greater than the deviation threshold, update the fuzzy rules of the cluster when the environmental stability is not greater than the stability threshold, and otherwise keep the rule base unchanged.

[0007] By monitoring environmental parameters in real time and combining them with fuzzy control output control commands, the system uses distribution deviation and environmental stability to determine whether rules need to be updated, thus ensuring that fuzzy rules are effective in the long term, avoiding the rigidity of the rule base, and enabling dynamic responses to environmental changes, making equipment management in smart parks more flexible and accurate.

[0008] Preferably, a historical environmental parameter corresponds to a set of control parameters and control errors. Clustering the historical environmental parameters includes: setting an initial number of clusters and using the K-prototype algorithm to cluster the historical environmental parameters to obtain an initial cluster; calculating the clustering cost of the initial number of clusters based on a cost function; adjusting the initial number of clusters multiple times to draw a clustering cost curve; taking the initial number of clusters corresponding to the inflection point of the curve as the target number of clusters; and the multiple clusters corresponding to the initial clusters of the target number of clusters. The cost function is the average distance from each historical environmental parameter in the initial cluster to the cluster center.

[0009] The K-prototype algorithm is used to cluster historical environmental parameters, and the target number of clusters is determined by the inflection point of the clustering cost curve. This avoids the subjectivity of manually setting the number of clusters and provides stable fuzzy rules for the subsequent construction of the rule base.

[0010] Preferably, the step of initializing each cluster as a fuzzy rule to construct a rule base includes: using the cluster center of the cluster as the antecedent of the fuzzy rule, and using the control parameter corresponding to the minimum control error in the cluster as the consequent of the fuzzy rule, to construct the fuzzy rule of the cluster.

[0011] The antecedent accurately describes the environmental state, while the consequent selects parameters that achieve the optimal control effect, ensuring that the initialized rule base has high practicality and reliability.

[0012] Preferably, the control parameters are obtained by weighted summation of the consequents of each fuzzy rule based on the activation degree of each fuzzy rule, thereby obtaining the control parameters.

[0013] Preferably, the long-term centroid satisfies the following relation:

[0014] ;

[0015] The short-run centroid satisfies the following relation:

[0016] ;

[0017] in, and For clusters At any moment and time long-term center of gravity, and For clusters At any moment and time Short-term centroid For a moment Belonging to the Real-time environmental parameters of each cluster. Real-time environmental parameters Normalized control error For long-term learning rate, It is the short-term learning rate and satisfies .

[0018] By setting different learning rates for updating, the long-term centroid and short-term centroid of each cluster are obtained. The long-term centroid is used to characterize the stable state of the cluster corresponding to the fuzzy rule in historical experience, while the short-term centroid is used to characterize the latest dynamics of the cluster corresponding to the fuzzy rule in recent observations. The distance between the two can reflect the applicability of the fuzzy rule to the current environment.

[0019] Preferably, clusters Distribution deviation Satisfying the relation:

[0020] ;in, and They are time points The long-term center of mass and the short-term center of mass, This represents the Euclidean distance.

[0021] Preferably, calculating the environmental stability of the cluster includes: obtaining the activation degree of the fuzzy rule corresponding to the cluster at each time point within a preset time period; the environmental stability is negatively correlated with the variance of the activation degree.

[0022] When the environmental input is stable, the activation level of fuzzy rules fluctuates little and has low variance, indicating a stable environment; conversely, large fluctuations in activation level indicate the presence of disturbances. This method can quantitatively distinguish between long-term changes and temporary fluctuations in the environment, thereby determining whether to update the fuzzy rules and avoiding frequent corrections due to short-term disturbances.

[0023] Preferably, the preset time period is from the moment when the distribution deviation of the cluster was greater than the deviation threshold in the last time to the current moment.

[0024] Preferably, updating the fuzzy rules of the cluster includes: calculating the Euclidean distance between the short-term centroids of the cluster and other clusters, and the similarity of the activation sequence; in response to any other cluster having an Euclidean distance less than a preset distance and a similarity greater than a similarity threshold, merging the cluster and the other cluster to obtain the fuzzy rules of the merged cluster; otherwise, clustering the environmental parameters in the cluster to obtain two sub-clusters, and obtaining the fuzzy rules of each sub-cluster.

[0025] Merging two clusters when their environmental states are highly similar can reduce redundant rules and improve the simplicity of the rule base; splitting a cluster when the differences within it are too great can refine the rules, improve control precision, and enable adaptive updates to the rule base.

[0026] Preferably, the environmental parameters include ambient brightness, weather data, and pedestrian traffic data captured by the camera; the control command is the brightness of the lighting equipment.

[0027] The technical solution of this application has the following beneficial technical effects:

[0028] By constructing a rule base, controlling a control unit, calculating a distribution deviation, and updating a rule base, a complete control closed loop is formed, enabling refined and intelligent linkage management of equipment within the smart park. By measuring the deviation between the long-term and short-term centroids, the effectiveness of fuzzy rules is measured, and combined with environmental stability judgment, long-term environmental changes and short-term disturbances can be accurately distinguished. This not only avoids rule oscillations caused by frequent invalid updates but also improves the accuracy of control parameters, ensuring the stability of the smart park's operation. Attached Figure Description

[0029] Figure 1 This is a structural block diagram of an intelligent linkage management system based on IoT technology according to an embodiment of this application.

[0030] Figure 2 This is a schematic diagram of the clustering cost curve according to an embodiment of this application.

[0031] Figure 3This is a schematic diagram of the distribution deviation of cluster 3 according to an embodiment of this application over time. Detailed Implementation

[0032] This application provides an intelligent linkage management system based on IoT technology. Figure 1 This is a structural block diagram of an intelligent linkage management system based on IoT technology according to an embodiment of this application. Figure 1 As shown, the IoT-based intelligent linkage management system includes a rule base construction unit, a control unit, a distribution deviation calculation unit, and a rule base update unit, which are described in detail below.

[0033] The rule base construction unit is used to cluster historical environmental parameters to obtain multiple clusters, and initialize each cluster as a fuzzy rule to build the rule base.

[0034] In one embodiment, it is first necessary to construct an initial rule base that reflects the basic operating rules of the park environment based on historical environmental parameters.

[0035] It should be noted that the smart park is equipped with numerous intelligent devices. In order to achieve intelligent linkage management, it is necessary to control each intelligent device. For ease of description, this application embodiment uses lighting equipment as an example. Therefore, the historical environmental parameters here are environmental parameters that affect the lighting equipment, so as to achieve precise control of the lighting equipment. That is, the environmental parameters include ambient brightness, weather data, and pedestrian flow data captured by the camera; the control command is the light brightness of the lighting equipment; and the pedestrian flow data is the number of people in the area illuminated by the lighting equipment per unit time.

[0036] In one embodiment, a historical environmental parameter corresponds to a set of control parameters and a control error. For example, at a certain historical moment, a set of inputs such as "ambient brightness, weather data, and pedestrian traffic data" would correspond to an optimal light brightness at that time and a control error used to characterize the difference between the actual illuminance and the desired illuminance.

[0037] Clustering historical environmental parameters includes: setting an initial number of clusters and using the K-prototype algorithm to cluster the historical environmental parameters to obtain initial clusters; calculating the clustering cost of the initial number of clusters based on a cost function; adjusting the initial number of clusters multiple times to plot a clustering cost curve; taking the initial number of clusters corresponding to the inflection point of the curve as the target number of clusters; and the multiple clusters corresponding to the initial clusters of the target number of clusters. The cost function is the average distance from each historical environmental parameter in the initial cluster to the cluster center.

[0038] It should be noted that since weather data is categorical (e.g., cloudy, sunny, foggy), while ambient brightness and pedestrian traffic data are numerical, the K-prototype algorithm, suitable for a mix of numerical and categorical data, is used to cluster historical environmental parameters. For any initial cluster, the environmental parameter composed of the mean of the numerical data and the mode of the categorical data is used as the cluster center. When calculating the average distance from each historical environmental parameter to the cluster center, Euclidean distance is calculated for numerical data, and Hamming distance is calculated for categorical data.

[0039] A good clustering result should have clusters corresponding to similar historical control scenarios (i.e., small control parameter variance) and ideal control performance (i.e., small average control error). By repeatedly adjusting the initial number of clusters (starting with K=2 and gradually increasing to K=20) and calculating the mean clustering index for each clustering result, a clustering cost curve can be plotted. Please refer to [link to relevant documentation]. Figure 2 This is a schematic diagram of the clustering cost curve according to an embodiment of this application. Typically, this curve has a distinct inflection point; after this point, increasing the number of clusters significantly reduces the contribution to reducing clustering costs. Therefore, the initial number of clusters corresponding to this inflection point is taken as the target number of clusters. The resulting clusters form the basis for subsequent rule base construction, such as... Figure 2 In the clustering cost curve shown, the inflection point corresponds to an initial cluster number of 5.

[0040] After obtaining the optimal clusters, each cluster needs to be initialized as a fuzzy rule to construct a rule base. Specifically, initializing each cluster as a fuzzy rule to construct a rule base includes: using the cluster center of the cluster as the antecedent of the fuzzy rule, and using the average value of the control parameters corresponding to each historical environmental parameter in the cluster as the consequent of the fuzzy rule, thus initializing the cluster as a fuzzy rule.

[0041] In this fuzzy rule, the antecedent is the IF part, and the consequent is the THEN part. For example, an initialized fuzzy rule is: "IF{ambient light is 50, weather is cloudy, pedestrian traffic is 5}THEN{light brightness is 85%}", where {50, cloudy, 5} are the cluster centers of the clusters.

[0042] In this way, multiple fuzzy rules are adaptively obtained using clustering algorithms, laying a data foundation for subsequent precise control and stable system operation.

[0043] The control unit is used to input real-time environmental parameters at any time into the fuzzy control algorithm to obtain the activation degree of each fuzzy rule in the rule base and obtain control parameters.

[0044] In one embodiment, once the rule base is built, control parameters at any time can be obtained based on the rule base, enabling real-time control of the lighting equipment.

[0045] Specifically, real-time environmental parameters are continuously collected at each moment and input into the fuzzy control algorithm. The fuzzy control algorithm first calculates the similarity between the real-time environmental parameter and the antecedent of each fuzzy rule in the rule base. The similarity is the activation degree of each fuzzy rule. The similarity is calculated using a similarity calculation method based on Euclidean distance.

[0046] After obtaining the activation level of each fuzzy rule, the control parameters are obtained by weighted summation of the consequents of each fuzzy rule based on their activation level. For example, if the real-time environmental parameters simultaneously activate rule A (consequence is 80% light brightness) and rule B (consequence is 60% light brightness), and their activation levels are 0.7 and 0.3 respectively, then the final output light brightness is (80%×0.7 + 60%×0.3) / (0.7 + 0.3) = 74%.

[0047] It should be noted that in traditional fuzzy control algorithms, the activation degree of each fuzzy rule is only used in the current control process and is then discarded; however, in this embodiment, the activation degree sequence of each fuzzy rule within a preset time period is cached to participate in the subsequent calculation of environmental stability and thus in the decision-making process of rule base update.

[0048] In this way, intelligent control based on fuzzy rules is realized based on the rule base. At the same time, the activation degree of each fuzzy rule during the control process is cached, providing the necessary data foundation for the subsequent adaptive update of the rule base.

[0049] The distribution deviation calculation unit is used to obtain the control error of real-time environmental parameters, calculate the long-term centroid and short-term centroid of the cluster to which the real-time environmental parameters belong based on the control error, and use the Euclidean distance between the long-term centroid and the short-term centroid as the distribution deviation of the cluster.

[0050] In one embodiment, the distribution deviation of clusters can be monitored in real time to see if the control effect of each fuzzy rule has shifted compared to historical experience. To accurately capture this shift, the concepts of long-term centroid and short-term centroid are introduced. The long-term centroid characterizes the stable state of the cluster corresponding to the fuzzy rule in historical experience, while the short-term centroid characterizes the latest dynamics of the cluster corresponding to the fuzzy rule in recent observations. When the two deviate significantly, it means that the fuzzy rule may no longer be suitable for the current environment, the fuzzy rule is outdated, and the fuzzy rule cannot achieve good control effect at the current moment.

[0051] For time Real-time environmental parameters ,in Indicates time The real-time environmental parameters belong to clusters. The time is obtained based on the existing rule base. The control parameters, and the control error after executing the control parameters; based on real-time environmental parameters. and control error for clustering The long-term and short-term centroids are updated. To reduce computational cost, a fixed time interval can be set to update the clusters. The long-term and short-term centroids are updated.

[0052] Specifically, the long-term centroid satisfies the following relation:

[0053] ;

[0054] The short-run centroid satisfies the following relation:

[0055] ;

[0056] in, and For clusters At any moment and time long-term center of gravity, and For clusters At any moment and time Short-term centroid For a moment Belonging to the Real-time environmental parameters of each cluster. Real-time environmental parameters Normalized control error For long-term learning rate, It is the short-term learning rate and satisfies .

[0057] Among them, the Min-Max method can be used to normalize the control error. First, the minimum error and the maximum error are preset based on the historical error range. After subtracting the minimum error from the control error, the result is divided by the difference between the maximum error and the minimum error to obtain the normalized control error.

[0058] Understandably, short-term learning rate Greater than the long-term learning rate This means that the short-term centroid is more sensitive to new data and the control error of new data, and can quickly reflect the latest dynamics; while the long-term centroid is more stable, representing the stable state of historical experience. If the environment undergoes long-term and permanent changes, real-time environmental parameters... It will continuously deviate from the original cluster center, leading to short-term centroids. Rapidly shifting away from the original cluster center, while the long-term centroid... It then deflects at a slow speed.

[0059] In addition, real-time environmental parameters Normalized control error As a weighting factor controlling the offset velocity of the long-term and short-term centroids, a small control error indicates that the real-time environmental parameters are relatively stable. To achieve better control, real-time environmental parameters should be added. The degree of influence on the center of mass (including long-term and short-term centers of mass) accelerates the updating and deviation of the center of mass, causing it to move in the direction of better control. For example, The value is 0.01. The value is 0.1.

[0060] After calculating the two centroids, at time... Cluster Distribution deviation Satisfying the relation:

[0061] ;

[0062] in, and They are time points The long-term center of mass and the short-term center of mass, Represents Euclidean distance. Distribution bias. It directly quantifies the distance between recent environmental data trends and long-term historical experience. The larger the value, the more the fuzzy rule can no longer adapt to the current environment, the less effective the fuzzy rule can be at the current moment, and the more necessary it is to update the rule base.

[0063] The distribution deviation of any cluster at any given time can be calculated, thereby enabling real-time monitoring of the cluster distribution deviation. When the distribution deviation exceeds a deviation threshold, the rule base update unit is triggered promptly to ensure that the fuzzy control algorithm outputs accurate control parameters. Please refer to [link to relevant documentation]. Figure 3 This is a schematic diagram of the distribution deviation of cluster 3 according to an embodiment of this application over time.

[0064] In this way, the effectiveness of each fuzzy rule can be accurately measured by the distribution deviation. When the distribution deviation exceeds the deviation threshold, it means that the fuzzy rule can no longer achieve a good control effect, and the rule base update unit is triggered in time.

[0065] The rule base update unit is used to calculate the environmental stability of the cluster when the distribution deviation of any cluster is greater than the deviation threshold, and update the fuzzy rules of the cluster when the environmental stability is not greater than the stability threshold; otherwise, the rule base remains unchanged.

[0066] In one embodiment, when the distribution deviation of any cluster exceeds a deviation threshold, it indicates that the fuzzy rule corresponding to that cluster is no longer suitable for the current environment. In this case, it is necessary to calculate the environmental stability of the cluster to determine whether the distribution deviation is caused by long-term environmental changes or temporary environmental disturbances. The environmental stability characterizes the severity of environmental changes within a preset time period. For example, areas near main roads are frequently subjected to temporary disturbances from vehicle headlights, resulting in low environmental stability; while areas affected by the shadows of new buildings experience continuous changes in illumination, leading to high environmental stability.

[0067] Specifically, calculating the environmental stability of the clusters includes: obtaining the activation level of the fuzzy rules corresponding to the clusters at each moment within a preset time period; the environmental stability is negatively correlated with the variance of the activation level. The preset time period is from the moment when the distribution deviation of the clusters last exceeded a deviation threshold to the current moment. The deviation threshold can be dynamically set based on the statistical distribution of historical distribution deviations; for example, the statistical distribution of historical distribution deviations includes the mean and standard deviation, and the deviation threshold is the mean plus twice the standard deviation.

[0068] Understandably, in a stable environment, the real-time environmental parameters input to the fuzzy control algorithm differ little, resulting in a relatively stable activation level of the fuzzy rules. Conversely, in an environment that is frequently subject to temporary disturbances, the activation level sequence of the fuzzy rules will experience drastic and irregular fluctuations, i.e., a large variance.

[0069] If the environmental stability is not greater than the stability threshold, it is determined that the distribution deviation is caused by long-term, permanent environmental changes, and the fuzzy rules of the cluster will be updated. Otherwise, it is determined to be a temporary environmental disturbance, and the rule base will be kept unchanged, thereby suppressing the update operation and avoiding rule oscillation.

[0070] In one embodiment, updating the fuzzy rules of the cluster includes: calculating the Euclidean distance between the short-term centroids of the cluster and other clusters, and the similarity of the activation sequence; in response to any other cluster having an Euclidean distance less than a preset distance and a similarity greater than a similarity threshold, merging the cluster and the other cluster to obtain the fuzzy rules of the merged cluster; otherwise, clustering the environmental parameters in the cluster to obtain two sub-clusters, and obtaining the fuzzy rules of each sub-cluster.

[0071] The similarity of the activation sequence can be calculated using a similarity method based on DTW distance, where the similarity is negatively correlated with the DTW distance of the activation sequence. Specifically, the similarity of the activation sequence is the ratio of 1 to (1 + DTW distance). The preset distance is 0.5, and the similarity threshold is 0.95.

[0072] When the Euclidean distance between any other cluster is less than a preset distance and the similarity is greater than a similarity threshold, it is considered that the environmental states described by the fuzzy rules corresponding to the two clusters tend to be the same. During the control process, the fuzzy rules corresponding to the two clusters will be activated at the same time and the activation degree is the same. At this time, the cluster and the other cluster are merged, and the fuzzy rules of the merged new cluster are obtained.

[0073] When there are no clusters that meet the merging conditions, a splitting operation is performed on the clusters, that is, the environmental parameters in the clusters are clustered again to obtain two sub-clusters, and new fuzzy rules are obtained for each sub-cluster, refining the original fuzzy rules into two more precise fuzzy rules.

[0074] Thus, by introducing environmental stability to distinguish between long-term, permanent environmental changes and temporary environmental disturbances, the rule base can be accurately and stably updated adaptively, avoiding the problem of rule oscillation and significantly improving the adaptability of the rule base and the control accuracy of the fuzzy control algorithm.

[0075] It should be noted that the scope of protection of this patent application shall be determined by the appended claims.

Claims

1. An intelligent linkage management system based on IoT technology, characterized by, The management system comprises: a rule base construction unit, configured to cluster historical environment parameters to obtain a plurality of clustering clusters, and initialize each clustering cluster as a fuzzy rule to construct a rule base, comprising: taking the clustering center of the clustering cluster as the antecedent of the fuzzy rule, and taking the control parameter corresponding to the minimum control error in the clustering cluster as the consequent of the fuzzy rule, to construct the fuzzy rule of the clustering cluster; a control unit, configured to input real-time environment parameters at any moment into a fuzzy control algorithm to obtain the activation degree of each fuzzy rule in the rule base, and obtain a control parameter; a distribution deviation calculation unit, configured to obtain the control error of the real-time environment parameters, calculate the long-term centroid and the short-term centroid of the clustering cluster to which the real-time environment parameters belong based on the control error, and take the Euclidean distance between the long-term centroid and the short-term centroid as the distribution deviation of the clustering cluster; a rule base updating unit, configured to, when the distribution deviation of any clustering cluster is greater than a deviation threshold, calculate the environmental stability of the clustering cluster, and in response to the environmental stability being not greater than a stability threshold, update the fuzzy rule of the clustering cluster, otherwise, keep the rule base unchanged; The environment parameters comprise environmental brightness, weather data, and people flow data captured by a camera; and the control parameter is the light brightness of a lighting device. 2.The IoT technology-based intelligent linkage management system according to claim 1, wherein One historical environment parameter corresponds to a group of control parameters and control errors, and clustering the historical environment parameters comprises: setting an initial clustering number, and clustering the historical environment parameters by using a K-prototype algorithm to obtain an initial cluster; calculating the clustering cost of the initial clustering number according to a cost function, adjusting the initial clustering number multiple times to draw a clustering cost curve, taking the initial clustering number corresponding to the inflection point of the curve as a target clustering number, and taking the plurality of clustering clusters to correspond to the initial cluster of the target clustering number; and the cost function is the average distance from each historical environment parameter in the initial cluster to the clustering center. 3.The IoT technology-based intelligent linkage management system according to claim 1, characterized in that, The control parameter is obtained by: weighting and summing the consequents of each fuzzy rule according to the activation degree of each fuzzy rule. 4.The IoT technology-based intelligent linkage management system according to claim 1, wherein The long-term centroid satisfies the relationship: ； The short-term centroid satisfies the relationship: ； wherein, and is a cluster at time and time , and is a cluster at time and time , is a real-time environmental parameter belonging to the th cluster at time is a control error after normalization processing, is a long-term learning rate, is a short-term learning rate, and satisfies . 5.The IoT technology-based intelligent linkage management system according to claim 1, wherein Cluster of the distribution bias satisfies the relationship: ; where, and are the long-term and short-term centroids, respectively, at time denotes the Euclidean distance.​ 6.The IoT technology-based intelligent linkage management system according to claim 1, wherein The environmental stability of the clustering cluster is calculated by: obtaining the activation degree of the fuzzy rule corresponding to the clustering cluster at each moment in a preset time period; and the environmental stability is negatively correlated with the variance of the activation degree. 7.The IoT technology-based intelligent linkage management system according to claim 6, wherein, The preset time period is from the moment when the distribution deviation of the clustering cluster is greater than the deviation threshold last time to the current moment. 8.The IoT technology-based intelligent linkage management system according to claim 1, wherein, The fuzzy rule of the clustering cluster is updated by: calculating the Euclidean distance between the short-term centroid of the clustering cluster and that of other clustering clusters, and the similarity of the activation degree sequence; in response to the Euclidean distance of any other clustering cluster being less than a preset distance and the similarity being greater than a similarity threshold, merging the clustering cluster and the other clustering cluster, and obtaining the fuzzy rule of the merged clustering cluster; otherwise, clustering the environment parameters in the clustering cluster to obtain two sub-clustering clusters, and obtaining the fuzzy rule of each sub-clustering cluster.

Images