Intelligent monitoring method and system based on Internet of Things

Abstract

The invention discloses an intelligent monitoring method and system based on the Internet of Things, and relates to the technical field of the Internet of Things, and the method comprises the steps: monitoring terminal equipment and edge node identification information through the Internet of Things, constructing an equipment state space and a scheduling motion space, and obtaining an edge node equipment state table and a reinforcement learning agent environment model; based on the edge node equipment state table and the reinforcement learning agent environment model, intelligent monitoring task scheduling is executed, a local monitoring data result set and a candidate forwarding edge node set which does not accept tasks are obtained, local training is carried out through an intelligent model, and a local model parameter gradient and a training error index are generated in combination with a differential privacy mechanism; and generating a local model parameter gradient and a training error index. By constructing the equipment state space and the scheduling action space, the dynamic optimization of the task scheduling strategy is realized, and the resource utilization rate and the task response efficiency of the system in a complex environment are improved.

Description

Technical Field

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

[0002] With the rapid development of IoT technology, an increasing number of intelligent terminal devices are being deployed in application scenarios such as industrial manufacturing, environmental monitoring, and smart cities. To improve the system's real-time responsiveness and data processing efficiency, edge computing architectures have become a research hotspot. In this architecture, edge nodes can perform preliminary processing and model training on locally collected data, then upload key parameters to the cloud or a coordination server for global aggregation, thereby enabling distributed collaborative learning. In recent years, federated learning has emerged as a distributed machine learning paradigm that protects user privacy.

[0003] However, existing intelligent monitoring systems based on the Internet of Things and federated learning still have problems such as low model update efficiency and lack of dynamic adaptability in task scheduling. The static weight distribution strategy is used for model aggregation, which does not consider the heterogeneity of resource status, data distribution and task load among different edge nodes, resulting in slow global model convergence. Existing methods usually rely on centralized scheduling mechanisms to complete task allocation, fail to fully utilize the local state information of edge nodes to build a reinforcement learning agent environment, and find it difficult to achieve adaptive optimization of task scheduling. Summary of the Invention

[0004] In view of the above existing problems, the present invention is proposed.

[0005] Therefore, the present invention provides an intelligent monitoring method based on the Internet of Things to solve the problems of low model update efficiency and lack of dynamic adaptability of task scheduling in the existing intelligent monitoring system combining the Internet of Things and federated learning.

[0006] In order to solve the above technical problems, the present invention provides the following technical solutions:

[0007] In a first aspect, the present invention provides an intelligent monitoring method based on the Internet of Things, which includes monitoring terminal devices and edge node identification information through the Internet of Things, constructing a device state space and a scheduling action space, and obtaining an edge node device state table and a reinforcement learning agent environment model;

[0008] Based on the edge node device status table and the reinforcement learning agent environment model, intelligent monitoring task scheduling is performed to obtain a set of local monitoring data results and a set of candidate forwarding edge nodes that have not accepted tasks. Local training is performed through the intelligent model, and the local model parameter gradient and training error index are generated by combining the differential privacy mechanism.

[0009] Upload the local model parameter gradients to the federated learning coordination server, perform global model aggregation updates on the local model parameter gradients based on the federated averaging algorithm, and obtain the global model parameters and update instructions;

[0010] Send the updated global model parameters to the edge nodes to obtain environmental judgment conclusions, monitoring result values and alarm marks;

[0011] Upload the environmental judgment conclusions, monitoring result values and alarm marks to the Internet of Things platform to trigger a response instruction set.

[0012] As a preferred solution of the intelligent monitoring method based on the Internet of Things described in the present invention, wherein: through the Internet of Things monitoring terminal equipment and edge node identification information, the device state space and scheduling action space are constructed, and the edge node device state table and reinforcement learning agent environment model are obtained, including the following steps:

[0013] Collect self-identification information, real-time status data and monitoring data through IoT monitoring terminal devices, and upload self-identification information and real-time status data to edge nodes;

[0014] Build a device state space based on its own identification information and real-time status data, analyze business goals and device performance indicators through monitoring data, and obtain task scheduling requirements;

[0015] Construct the scheduling action space through the device state space and task scheduling requirements;

[0016] Combine the device state space and the scheduling action space to form a reinforcement learning agent environment model.

[0017] Based on the device state space and the reinforcement learning agent environment model, the edge node device state table is obtained.

[0018] As a preferred solution of the intelligent monitoring method based on the Internet of Things described in the present invention, wherein: based on the edge node device state table and the reinforcement learning agent environment model, intelligent monitoring task scheduling is performed to obtain a local monitoring data result set and a set of candidate forwarding edge nodes that have not accepted the task, including the following steps:

[0019] The edge node calculates the optimal scheduling action under the current state based on the device state table and the reinforcement learning environment model;

[0020] The edge node allocates local intelligent monitoring tasks according to the optimal scheduling action and collects local monitoring data results;

[0021] The edge node selects a set of candidate forwarding edge nodes that have not accepted any tasks and meet the resource and performance requirements.

[0022] As a preferred solution of the intelligent monitoring method based on the Internet of Things described in the present invention, wherein: local training is performed through an intelligent model, and a local model parameter gradient and a training error index are generated in combination with a differential privacy mechanism, and the local model parameter gradient and the training error index are generated, including the following steps:

[0023] Based on the local monitoring data result set, the local model is iteratively trained using the stochastic gradient descent algorithm with momentum to obtain the local model parameter gradient;

[0024] Differential privacy protection is performed by adding Gaussian noise to the local model parameter gradient to generate differential privacy gradients;

[0025] Use differential privacy gradients to load local model parameters and iterate training on the local model;

[0026] Calculate the training error metric based on the trained local model.

[0027] As a preferred solution of the IoT-based intelligent monitoring method described in the present invention, the local model parameter gradients are uploaded to the federated learning coordination server, and the global model aggregation update is performed on the local model parameter gradients based on the federated averaging algorithm to obtain the global model parameters and update instructions, including the following steps:

[0028] Upload the differentially private gradients generated by local model parameter gradients to the federated learning coordination server through a secure communication protocol;

[0029] The federated averaging algorithm is used on the federated learning coordination server to perform global model aggregation updates on all received differentially private gradients to obtain global model parameters and update instructions.

[0030] As a preferred solution of the intelligent monitoring method based on the Internet of Things described in the present invention, the updated global model parameters are sent to the edge node to obtain the environmental judgment conclusion, monitoring result value and alarm mark, including the following steps:

[0031] Package the global model parameters and update instructions into a structured update package, and send the structured update package to the edge node using a TLS encrypted channel;

[0032] Use the updated global model parameters to reason about the monitoring data and generate original reasoning results;

[0033] Analyze the original reasoning results based on business logic rules to obtain environmental judgment conclusions and monitoring result values;

[0034] Conduct statistical analysis on monitoring data to set alarm thresholds;

[0035] When the monitoring result value is greater than the alarm threshold, an alarm mark is generated.

[0036] As a preferred solution of the intelligent monitoring method based on the Internet of Things described in the present invention, the environmental judgment conclusion, monitoring result value and alarm mark are uploaded to the Internet of Things platform to trigger a response instruction set, including the following steps:

[0037] The edge node packages the environmental judgment conclusion, monitoring result value and alarm mark into a structured JSON data format;

[0038] Upload structured data packets to the designated interface of the IoT platform using HTTPS and mutual TLS authentication;

[0039] Obtain device status cache based on structured data uploaded to the IoT platform interface;

[0040] The IoT platform obtains a set of response instructions based on the judgment of the alarm mark.

[0041] In a second aspect, the present invention provides an intelligent monitoring system based on the Internet of Things, including an edge node management module, which monitors terminal devices and edge node identification information through the Internet of Things, constructs a device state space and a scheduling action space, and obtains an edge node device state table and a reinforcement learning agent environment model;

[0042] The execution module performs intelligent monitoring task scheduling based on the edge node device status table and the reinforcement learning agent environment model, obtains the local monitoring data result set and the set of candidate forwarding edge nodes that have not accepted the task, performs local training through the intelligent model, and generates local model parameter gradients and training error indicators in combination with the differential privacy mechanism;

[0043] The federated learning model collaborative training module uploads the local model parameter gradients to the federated learning coordination server, performs a global model aggregation update on the local model parameter gradients based on the federated averaging algorithm, and obtains the global model parameters and update instructions;

[0044] The environmental judgment module sends the updated global model parameters to the edge nodes to obtain the environmental judgment conclusion, monitoring result value and alarm mark;

[0045] The response control module uploads the environmental judgment conclusion, monitoring result value and alarm mark to the Internet of Things platform to trigger the response instruction set.

[0046] In a third aspect, the present invention provides a computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: when the computer program is executed by the processor, any step of the intelligent monitoring method based on the Internet of Things as described in the first aspect of the present invention is implemented.

[0047] In a fourth aspect, the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein: when the computer program is executed by a processor, any step of the intelligent monitoring method based on the Internet of Things as described in the first aspect of the present invention is implemented.

[0048] The beneficial effects of the present invention are as follows: by constructing the device state space and scheduling action space, dynamic optimization of the task scheduling strategy is realized, the resource utilization and task response efficiency of the system in complex environments are improved, and a high-quality and reasonably distributed local data foundation is provided for subsequent model training. By introducing the weighted federal averaging algorithm to aggregate the local model parameter gradients after differential privacy protection, efficient collaborative learning is achieved while ensuring data privacy, the convergence speed and generalization ability of the global model are improved, the intelligence level and collaborative efficiency of the system are effectively enhanced, and it is suitable for the actual deployment needs of intelligent monitoring of the Internet of Things in multi-source heterogeneous environments. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0050] Figure 1 Flowchart of the intelligent monitoring method based on the Internet of Things.

[0051] Figure 2 Schematic diagram of the edge node device status table.

[0052] Figure 3 Flowchart of training error indicators.

[0053] Figure 4 This is a flow chart of the intelligent monitoring module based on the Internet of Things. DETAILED DESCRIPTION

[0054] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the specific embodiments of the present invention are described in detail below with reference to the accompanying drawings.

[0055] In the following description, many specific details are set forth to facilitate a full understanding of the present invention. However, the present invention may also be implemented in other ways different from those described herein. Those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0056] Secondly, the term "one embodiment" or "embodiment" herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in various places throughout this specification does not necessarily refer to the same embodiment, nor does it refer to a separate or selective embodiment that is mutually exclusive of other embodiments.

[0057] Reference Figures 1 to 4 , is an embodiment of the present invention, which provides an intelligent monitoring method and system based on the Internet of Things, including the following steps:

[0058] Through the Internet of Things monitoring terminal devices and edge node identification information, the device state space and scheduling action space are constructed to obtain the edge node device state table and reinforcement learning agent environment model.

[0059] The IoT monitoring terminal device collects its own identification information, real-time status data and monitoring data, and uploads its own identification information and real-time status data to the edge node.

[0060] Furthermore, the IoT monitoring terminal device collects its own identification information, real-time status data and monitoring data, including the device's unique identification number, current operating status parameters and environmental perception data; the collected self-identification information and real-time status data are uploaded to the corresponding edge node through the wireless communication protocol; the device's unique identification number is used to distinguish different IoT monitoring terminal devices, the current operating status parameters include but are not limited to device power, computing resource usage and network connection quality, and the environmental perception data is the target parameter that needs to be monitored in the physical environment where the device is located; an encrypted transmission mechanism is used during the upload process to ensure data security, and the TLS1.3 protocol is used for data encapsulation and transmission. After the upload is completed, the edge node receives and parses the received data content

[0061] The device status space is constructed based on its own identification information and real-time status data. The business goals and device performance indicators are analyzed through monitoring data to obtain task scheduling requirements.

[0062] Furthermore, based on the self-identification information and real-time status data collected by the IoT monitoring terminal devices, the device's unique identity number, current operating status parameters and environmental perception data are classified and encoded according to a preset data structure to form a state feature set that can be used for task scheduling decisions. The current operating status parameters include device power, resource usage and network connection quality. The environmental perception data is the target parameter that needs to be monitored in the physical environment where the device is located. By analyzing the content type and business application scenario of the environmental perception data, the corresponding business goals are determined, and real-time warnings are issued for abnormal temperature changes or periodic evaluations are performed on air quality indicators. Combined with the device operating status parameters and business goal requirements, equipment performance indicators, including data processing capability score, task response delay level and communication stability index, task scheduling requirements are generated according to business goals and equipment performance indicators.

[0063] The scheduling action space is constructed through the device state space and task scheduling requirements.

[0064] Furthermore, the set of candidate actions that can be used to execute task scheduling is determined through the state feature set in the device state space and the task type, priority and required resource conditions described in the task scheduling requirements; wherein, the state feature set includes the device unique identity number, current operating state parameters and environmental perception data of the IoT monitoring terminal device, and the task scheduling requirements clearly define the communication bandwidth and response time requirements for task execution. Based on whether the current operating state parameters of the device meet the task execution conditions, the scheduling actions applicable to the current state are screened out, such as assigning local execution, forwarding to other devices for execution or temporarily not executing, and all eligible scheduling actions are organized into a scheduling action space.

[0065] The device state space and scheduling action space are combined to form a reinforcement learning agent environment model.

[0066] Furthermore, the state feature set in the device state space is mapped to the candidate scheduling actions in the scheduling action space to establish an association between the state and the action. Based on the device unique identity number, current operating state parameters and environmental perception data of the IoT monitoring terminal device, combined with the task type, priority and required resource conditions described in the task scheduling requirements, the state transition rules are defined to describe the new state to be transferred to after executing a certain scheduling action in a specific state. The reward function is set according to the business goal. For example, the instant reward value of the scheduling strategy is calculated based on the task completion efficiency, resource consumption and communication stability. By integrating the state space, action space, state transition rules and reward function, a reinforcement learning agent environment model is formed.

[0067] Based on the device state space and the reinforcement learning agent environment model, the edge node device state table is obtained.

[0068] Furthermore, based on the state feature set in the device state space and the state transition rules in the reinforcement learning agent environment model, the operating status of the IoT monitoring terminal devices under different scheduling actions is recorded and summarized. The state feature set includes the device's unique identification number, current operating status parameters, and environmental perception data. The state transition rules describe the new state to which a scheduling action is transferred after executing a certain scheduling action in a specific state. By periodically collecting the device operating status parameters and combining them with the task type, priority, and required resource conditions in the task scheduling requirements, entries containing device identification, current status, available scheduling actions, and corresponding reward values are generated, and all entries are organized into an edge node device state table according to a preset data format.

[0069] Based on the edge node device status table and the reinforcement learning agent environment model, intelligent monitoring task scheduling is performed to obtain a set of local monitoring data results and a set of candidate forwarding edge nodes that have not accepted tasks.

[0070] The edge node calculates the optimal scheduling action under the current state based on the device state table and the reinforcement learning environment model.

[0071] Specifically, the expression is,

[0072]

[0073] Among them, A * is the optimal scheduling action under the current state, s is the current device state, a is the scheduling action, θ is the neural network parameter, Q is the depth Q network, and A is the total number of scheduling actions.

[0074] The edge node allocates local intelligent monitoring tasks according to the optimal scheduling action and collects local monitoring data results.

[0075] Furthermore, the edge node allocates local intelligent monitoring tasks according to task type, priority and required resource conditions based on the optimal scheduling action obtained from the current device status and the reinforcement learning agent environment model. After completing the task allocation, it calls locally deployed sensors or data acquisition devices to collect target parameters in the physical environment in real time to generate local monitoring data results.

[0076] The edge node selects a set of candidate forwarding edge nodes that have not accepted any tasks and meet the resource and performance requirements.

[0077] Furthermore, the edge node screens out other edge nodes that have not accepted local intelligent monitoring tasks based on the device identification, available scheduling actions and corresponding reward values recorded in the current device status table. Combined with the resource conditions and performance indicator requirements described in the task scheduling requirements, the screened edge nodes are evaluated one by one, and the edge nodes that meet the resource and performance requirements are included in the candidate forwarding edge node set.

[0078] Local training is performed through the intelligent model, and the local model parameter gradient and training error index are generated in combination with the differential privacy mechanism.

[0079] Based on the local monitoring data results, the stochastic gradient descent algorithm with momentum is used to iteratively train the local model to obtain the local model parameter gradient.

[0080] Furthermore, based on the numerical distribution characteristics of the local model parameter gradient, random noise that conforms to the normal distribution is independently added to each dimension of the local model parameter gradient according to the privacy budget parameter and the noise scaling factor to generate a differential privacy gradient that meets the differential privacy requirements.

[0081] Differential privacy protection is performed by adding Gaussian noise to the local model parameter gradient to generate differential privacy gradients.

[0082] Furthermore, based on the numerical distribution of the local model parameter gradient, according to the privacy budget and noise scale parameters, Gaussian noise with a mean of zero and a preset standard deviation is added to each parameter dimension of the local model parameter gradient, thereby generating a differential privacy gradient that has been processed with differential privacy protection.

[0083] Use differentially private gradients to load local model parameters and iterate training on the local model.

[0084] Furthermore, the current model parameters of the local intelligent model are updated using the differential privacy gradient that has been processed with differential privacy protection, and the perturbed model parameters are loaded into the local intelligent model as the initial parameters for the next round of training iteration. During each training iteration, the sample data in the local monitoring data result set is forward propagated based on the updated local intelligent model, and backpropagation optimization is performed based on the differential privacy gradient to continuously adjust the model parameters of the local intelligent model.

[0085] Calculate the training error metric based on the trained local model.

[0086] Specifically, the expression is,

[0087]

[0088] Among them, K is the training error index, V is the total number of samples, and yb is the true label of the b-th sample, is the predicted output value of the b-th sample, and b is the sample index.

[0089] The local model parameter gradients are uploaded to the federated learning coordination server, and the global model aggregation update is performed on the local model parameter gradients based on the federated averaging algorithm to obtain the global model parameters and update instructions.

[0090] The differentially private gradients generated by local model parameter gradients are uploaded to the federated learning coordination server through a secure communication protocol.

[0091] Furthermore, by adopting a secure communication protocol, such as the Transport Layer Security (TLS) protocol, the local model parameter gradient (i.e., differential privacy gradient) that has been processed with differential privacy protection is encrypted and transmitted from the edge node to the federated learning coordination server. The edge node first packages the differential privacy gradient to ensure that the data format meets the upload requirements, and uses the encrypted channel provided by the TLS protocol to ensure the confidentiality and integrity of the data during transmission, preventing potential information leakage and tampering risks. After receiving the encrypted data packet, the federated learning coordination server decrypts and parses it to extract the differential privacy gradient.

[0092] On the federated learning coordination server, the federated averaging algorithm is used to aggregate and update the global model for all received differentially private gradients to obtain global model parameters and update instructions.

[0093] Specifically, the expression is.

[0094]

[0095] Among them, α i is the weight coefficient of the i-th edge node in federated learning, D j is the number of local training data samples used by the jth edge node in the current training round, D i is the number of data samples used for local training of the i-th edge node, n is the total number of edge nodes participating in federated learning in the current round, j is the participating node, and i is the number of the edge node.

[0096] The updated global model parameters are sent to the edge nodes to obtain environmental judgment conclusions, monitoring result values and alarm marks.

[0097] The global model parameters and update instructions are packaged into a structured update package, and the structured update package is sent to the edge node using a TLS encrypted channel.

[0098] Furthermore, the federated learning coordination server serializes the updated global model parameters and the corresponding update instructions in a preset data format to generate a structured update package. The global model parameters include the intelligent model parameters after aggregation and optimization by the federated averaging algorithm. The update instructions include the model version identifier, update timestamp, and applicable task type metadata. After completing the construction of the structured update package, the federated learning coordination server transmits the data packet to the target edge node through the TLS encrypted channel.

[0099] The updated global model parameters are used to reason about the monitoring data and generate original inference results.

[0100] Furthermore, the edge node loads the updated global model parameters into the local intelligent model, performs forward propagation calculations on the collected monitoring data, and generates corresponding original inference results; among them, the monitoring data includes environmental perception data reported by the IoT monitoring terminal device, and the original inference results represent the output value of the local intelligent model under the current global model parameters, which is used for the subsequent generation of environmental judgment conclusions, monitoring result values and alarm marks. During the inference process, the local intelligent model makes predictions based on the trained network structure and updated model parameters.

[0101] The original reasoning results are analyzed based on business logic rules to obtain environmental judgment conclusions and monitoring result values.

[0102] Furthermore, based on the original reasoning results and combined with the preset business logic rules, the state of the environment in which the IoT monitoring terminal device is located is judged and classified to generate environmental judgment conclusions and specific monitoring result values.

[0103] Conduct statistical analysis on monitoring data and set alarm thresholds.

[0104] Furthermore, the specific process of conducting statistical analysis on monitoring data to set alarm thresholds includes: based on the historical monitoring data result set and the current monitoring result value, using statistical methods to calculate the data distribution characteristics, such as the mean, standard deviation or percentile, and setting reasonable alarm thresholds in combination with business scenario requirements; wherein, the alarm threshold is used to distinguish the boundary between the normal state and the abnormal state. For example, a value exceeding the data mean plus twice the standard deviation is set as the alarm trigger condition.

[0105] When the monitoring result value is greater than the alarm threshold, an alarm mark is generated.

[0106] Furthermore, when the monitoring result value exceeds the alarm threshold set based on historical data and statistical analysis, the alarm mark generation operation is triggered, and the current monitoring result value is compared with the corresponding alarm threshold. If the monitoring result value is greater than the alarm threshold, an alarm mark is generated. The alarm mark is used to indicate that the current environmental status is abnormal or exceeds the preset safety range.

[0107] Upload the environmental judgment conclusions, monitoring result values and alarm marks to the Internet of Things platform to trigger a response instruction set.

[0108] The edge node packages the environmental judgment conclusion, monitoring result value and alarm mark into a structured JSON data format.

[0109] Furthermore, the edge node integrates the generated environmental judgment conclusions, monitoring result values and alarm tags according to the preset data structure and encapsulates them in the structured JSON data format; among them, the environmental judgment conclusions are used to describe the classification results of the current physical environment status, the monitoring result values are specific numerical outputs, and the alarm tags are used to indicate whether the alarm conditions are triggered.

[0110] Use HTTPS and two-way TLS authentication to upload structured data packets to the specified interface of the IoT platform.

[0111] Furthermore, the edge node establishes a secure communication channel with the IoT platform through the HTTPS protocol, and enables the two-way TLS authentication mechanism during the transmission process to verify the legitimacy of the identity of the edge node and the IoT platform; after completing the identity authentication, the encapsulated structured JSON data packet is sent to the designated network interface of the IoT platform according to the preset interface specifications.

[0112] The device status cache is obtained based on the structured data uploaded to the IoT platform interface.

[0113] Furthermore, after receiving the structured JSON data packet uploaded by the edge node through HTTPS and two-way TLS authentication, the IoT platform parses the data packet content, extracts the environmental judgment conclusions, monitoring result values and alarm marks, and stores the extracted information in the device status cache according to the preset data structure.

[0114] The IoT platform obtains a set of response instructions based on the judgment of the alarm mark.

[0115] Furthermore, the IoT platform obtains the alarm mark in the device status cache and makes a judgment based on whether the alarm mark exists and its type. If the alarm mark is in a valid state, the corresponding response instruction set is matched according to the preset alarm response strategy; wherein, the response instruction set includes but is not limited to remote control instructions, alarm notification messages, data acquisition frequency adjustment commands or task scheduling parameter optimization solutions, which are used to quickly respond to and handle abnormal situations; for example, in a temperature monitoring scenario, when the alarm mark indicates that the temperature exceeds the standard, the IoT platform generates a response instruction set to start the cooling equipment, send an alarm SMS and increase the sampling frequency.

[0116] This embodiment also provides an intelligent monitoring system based on the Internet of Things, including: an edge node management module, which monitors terminal devices and edge node identification information through the Internet of Things, constructs a device state space and a scheduling action space, and obtains an edge node device state table and a reinforcement learning agent environment model;

[0117] The execution module performs intelligent monitoring task scheduling based on the edge node device status table and the reinforcement learning agent environment model, obtains the local monitoring data result set and the set of candidate forwarding edge nodes that have not accepted the task, performs local training through the intelligent model, and generates local model parameter gradients and training error indicators in combination with the differential privacy mechanism;

[0118] The federated learning model collaborative training module uploads the local model parameter gradients to the federated learning coordination server, performs a global model aggregation update on the local model parameter gradients based on the federated averaging algorithm, and obtains the global model parameters and update instructions;

[0119] The environmental judgment module sends the updated global model parameters to the edge nodes to obtain the environmental judgment conclusion, monitoring result value and alarm mark;

[0120] The response control module uploads the environmental judgment conclusion, monitoring result value and alarm mark to the Internet of Things platform to trigger the response instruction set.

[0121] This embodiment also provides a computer device suitable for the case of an intelligent monitoring method based on the Internet of Things, including: a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute computer-executable instructions to implement the intelligent monitoring method based on the Internet of Things proposed in the above embodiment.

[0122] The computer device may be a terminal, comprising a processor, a memory, a communication interface, a display screen and an input device connected via a system bus. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device comprises a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The communication interface of the computer device is used to communicate with an external terminal in a wired or wireless manner, and the wireless manner may be achieved through WIFI, an operator network, NFC (near field communication) or other technologies. The display screen of the computer device may be a liquid crystal display or an electronic ink display screen, and the input device of the computer device may be a touch layer covering the display screen, or a button, trackball or touchpad provided on the housing of the computer device, or an external keyboard, touchpad or mouse.

[0123] This embodiment also provides a storage medium having a computer program stored thereon, which, when executed by a processor, implements the intelligent monitoring method based on the Internet of Things proposed in the above embodiment; the storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic memory, flash memory, magnetic disk or optical disk.

[0124] In summary, the present invention realizes the dynamic optimization of task scheduling strategies by constructing device state space and scheduling action space, improves the resource utilization and task response efficiency of the system in complex environments, and provides a high-quality and reasonably distributed local data foundation for subsequent model training. By introducing the weighted federal averaging algorithm to aggregate the local model parameter gradients after differential privacy protection, efficient collaborative learning is achieved while ensuring data privacy, and the convergence speed and generalization ability of the global model are improved, effectively enhancing the intelligence level and collaborative efficiency of the system, which is suitable for the actual deployment needs of intelligent monitoring of the Internet of Things in multi-source heterogeneous environments.

[0125] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. An intelligent monitoring method based on the Internet of Things, characterized by: include, By monitoring the terminal devices and edge node identification information through the Internet of Things, the device state space and scheduling action space are constructed to obtain the edge node device state table and reinforcement learning agent environment model; Based on the edge node device status table and the reinforcement learning agent environment model, intelligent monitoring task scheduling is performed to obtain a set of local monitoring data results and a set of candidate forwarding edge nodes that have not accepted tasks. Local training is performed through the intelligent model, and the local model parameter gradient and training error index are generated by combining the differential privacy mechanism. Upload the local model parameter gradients to the federated learning coordination server, perform global model aggregation updates on the local model parameter gradients based on the federated averaging algorithm, and obtain the global model parameters and update instructions; Send the updated global model parameters to the edge nodes to obtain environmental judgment conclusions, monitoring result values and alarm marks; Upload the environmental judgment conclusions, monitoring result values and alarm marks to the Internet of Things platform to trigger a response instruction set.

2. The intelligent monitoring method based on the Internet of Things according to claim 1, characterized in that: By monitoring the terminal devices and edge node identification information through the Internet of Things, the device state space and scheduling action space are constructed to obtain the edge node device state table and reinforcement learning agent environment model, including the following steps: Collect self-identification information, real-time status data and monitoring data through IoT monitoring terminal devices, and upload self-identification information and real-time status data to edge nodes; Build a device state space based on its own identification information and real-time status data, analyze business goals and device performance indicators through monitoring data, and obtain task scheduling requirements; Construct the scheduling action space through the device state space and task scheduling requirements; Combine the device state space and the scheduling action space to form a reinforcement learning agent environment model; Based on the device state space and the reinforcement learning agent environment model, the edge node device state table is obtained.

3. The intelligent monitoring method based on the Internet of Things according to claim 2, characterized in that: Based on the edge node device status table and the reinforcement learning agent environment model, intelligent monitoring task scheduling is performed to obtain a set of local monitoring data results and a set of candidate forwarding edge nodes that have not accepted the task, including the following steps: The edge node calculates the optimal scheduling action under the current state based on the device state table and the reinforcement learning environment model; The edge node allocates local intelligent monitoring tasks according to the optimal scheduling action and collects local monitoring data results; The edge node selects a set of candidate forwarding edge nodes that have not accepted any tasks and meet the resource and performance requirements.

4. The intelligent monitoring method based on the Internet of Things according to claim 3, characterized in that: Local training is performed through the intelligent model, and the differential privacy mechanism is combined to generate local model parameter gradients and training error indicators. The following steps are included: Based on the local monitoring data result set, the local model is iteratively trained using the stochastic gradient descent algorithm with momentum to obtain the local model parameter gradient; Differential privacy protection is performed by adding Gaussian noise to the local model parameter gradient to generate differential privacy gradients; Use differential privacy gradients to load local model parameters and iterate training on the local model; Calculate the training error metric based on the trained local model.

5. The intelligent monitoring method based on the Internet of Things according to claim 4, characterized in that: Upload the local model parameter gradients to the federated learning coordination server, perform global model aggregation updates on the local model parameter gradients based on the federated averaging algorithm, and obtain the global model parameters and update instructions, including the following steps: Upload the differentially private gradients generated by local model parameter gradients to the federated learning coordination server through a secure communication protocol; The federated averaging algorithm is used on the federated learning coordination server to perform global model aggregation updates on all received differentially private gradients to obtain global model parameters and update instructions.

6. The intelligent monitoring method based on the Internet of Things according to claim 5, characterized in that: Send the updated global model parameters to the edge nodes to obtain environmental judgment conclusions, monitoring result values and alarm marks. The following steps are included: Package the global model parameters and update instructions into a structured update package, and send the structured update package to the edge node using a TLS encrypted channel; Use the updated global model parameters to reason about the monitoring data and generate original reasoning results; Analyze the original reasoning results based on business logic rules to obtain environmental judgment conclusions and monitoring result values; Conduct statistical analysis on monitoring data to set alarm thresholds; When the monitoring result value is greater than the alarm threshold, an alarm mark is generated.

7. The intelligent monitoring method based on the Internet of Things according to claim 6, characterized in that: Upload the environmental judgment conclusion, monitoring result value and alarm mark to the Internet of Things platform to trigger the response instruction set, including the following steps: The edge node packages the environmental judgment conclusion, monitoring result value and alarm mark into a structured JSON data format; Upload structured data packets to the designated interface of the IoT platform using HTTPS and mutual TLS authentication; Obtain device status cache based on structured data uploaded to the IoT platform interface; The IoT platform obtains a set of response instructions based on the judgment of the alarm mark.

8. An Internet of Things-based intelligent monitoring system, based on the Internet of Things-based intelligent monitoring method according to any one of claims 1 to 7, characterized in that: include, The edge node management module monitors the terminal devices and edge node identification information through the Internet of Things, constructs the device state space and scheduling action space, and obtains the edge node device state table and reinforcement learning agent environment model; The execution module performs intelligent monitoring task scheduling based on the edge node device status table and the reinforcement learning agent environment model, obtains the local monitoring data result set and the set of candidate forwarding edge nodes that have not accepted the task, performs local training through the intelligent model, and generates local model parameter gradients and training error indicators in combination with the differential privacy mechanism; The federated learning model collaborative training module uploads the local model parameter gradients to the federated learning coordination server, performs a global model aggregation update on the local model parameter gradients based on the federated averaging algorithm, and obtains the global model parameters and update instructions; The environmental judgment module sends the updated global model parameters to the edge nodes to obtain the environmental judgment conclusion, monitoring result value and alarm mark; The response control module uploads the environmental judgment conclusion, monitoring result value and alarm mark to the Internet of Things platform to trigger the response instruction set.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the intelligent monitoring method based on the Internet of Things according to any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the intelligent monitoring method based on the Internet of Things according to any one of claims 1 to 7 are implemented.

Images