Multi-dimensional Monitoring and Disposal Method, Device and Platform for Power Internet of Things Based on Cloud-Edge Collaboration

Abstract

The present invention discloses a multi-dimensional monitoring and disposal method for a power Internet of Things based on cloud-edge collaboration. Based on the power Internet of Things system with a three-layer architecture of "cloud-edge-end", it conducts intelligent monitoring and fault disposal of edge devices in the power Internet of Things. Among the edge devices, log information during the operation of the Internet of Things system is collected, and based on artificial intelligence algorithms, the fault types of abnormal log information are analyzed in the edge devices, and corresponding disposal strategies are executed to achieve edge autonomy; in the Internet of Things management platform, the fault types of abnormal log information are analyzed, and corresponding disposal strategies are sent to the edge disposal module to achieve cloud-edge collaborative governance. The present invention realizes multi-dimensional monitoring and edge autonomy through the collection and analysis of various types of log information, and based on cloud-edge collaboration, it realizes precise and intelligent analysis and rapid disposal of faults, effectively improving the operation and maintenance efficiency.

Description

Technical Field

[0001] The present invention belongs to the technical field of edge computing in the power Internet of Things, and specifically relates to a multi-dimensional monitoring and disposal method for the power Internet of Things based on cloud-edge collaboration, and also relates to an edge device and an Internet of Things management platform. Background Art

[0002] The power Internet of Things mainly applies a three-layer architecture of "cloud-edge-end". Among them, the "end" is the terminal layer composed of acquisition and sensing terminals located in various power business scenarios, the "cloud" is the Internet of Things management platform located in the cloud, and the "edge" is the edge device located between the business terminal and the cloud master station, which is used to connect the terminal layer and the management platform to realize functions such as data acquisition, secure access, and edge computing. The edge device of the power Internet of Things adopts lightweight Docker container technology, and realizes the access and data acquisition of power terminals in different scenarios through various business APPs running in containers.

[0003] With the continuous advancement of the construction of the power intelligent Internet of Things system, the number of access scenarios and access terminals in each power specialty is increasing continuously, which puts forward higher requirements for the operation stability of access devices and the timeliness of abnormal disposal. The traditional centralized and manual fault analysis and disposal of the cloud management platform can no longer adapt to the characteristics of high transmission cost, large network delay, and rapid analysis and disposal in the power Internet of Things. When an abnormal phenomenon occurs on-site, the traditional method cannot quickly analyze and dispose of problems through the management platform and clarify the attribution of operation and maintenance responsibilities, resulting in low operation and maintenance work efficiency and consuming a large amount of labor and time costs. Summary of the Invention

[0004] The purpose of the present invention is to overcome the deficiencies in the prior art, and provide a multi-dimensional monitoring and disposal method for the power Internet of Things based on cloud-edge collaboration, in which the edge device close to the "end" layer device and the Internet of Things management platform in the "cloud" layer cooperate remotely to perform rapid and accurate intelligent network fault analysis and disposal, and provide edge intelligent autonomy ability to solve the problems of high remote operation and maintenance delay, slow disposal, and low efficiency.

[0005] To solve the above technical problems, the technical solution of the present invention is as follows.

[0006] In a first aspect, the present invention provides a multi-dimensional monitoring and disposal method for the power Internet of Things based on cloud-edge collaboration, which is executed in an edge device and includes the following processes:

[0007] Collect the operation log information of the "edge" layer in the power Internet of Things;

[0008] Screen out the abnormal log information from the operation log information, perform a fault judgment on the abnormal log information to obtain the corresponding fault type, and execute the disposal strategy corresponding to the fault type. If the execution of the disposal strategy fails, send this abnormal log information to the Internet of Things management platform.

[0009] Optionally, the collection frequency is at the minute level.

[0010] Optionally, the operation log information includes the operation status information of edge devices, edge management software, containers, and business application APPs, where:

[0011] The operation information of edge devices specifically includes: CPU usage rate, memory usage rate, disk space usage rate, device operation duration, and whether the device has abnormal restarts;

[0012] The operation information of edge management software specifically includes: whether the software has abnormal restarts;

[0013] The operation information of containers specifically includes: CPU usage rate, memory usage rate, disk usage rate, and whether the container has abnormal restarts;

[0014] The operation information of business application APPs specifically includes: CPU usage rate, memory usage rate, disk usage rate, whether the application has abnormal restarts, business data parsing situation, and end device offline situation.

[0015] Optionally, the failure judgment of the abnormal log information includes:

[0016] Using a failure judgment model trained based on the random forest algorithm to perform failure judgment on the abnormal log information.

[0017] Optionally, the training data of the failure judgment model is the log information corresponding to high-frequency abnormal situations and the failure types corresponding to the abnormalities.

[0018] Optionally, the judgment process for the failure of the disposal strategy execution includes:

[0019] If the same abnormal log information is repeatedly received within a specified time after the disposal strategy is executed, it is judged that the execution of the disposal strategy fails.

[0020] Optionally, it further includes: reporting the successful execution information of the disposal strategy to the physical management platform.

[0021] In a second aspect, the present invention also provides an edge device, including:

[0022] An edge monitoring module that collects the operation log information of the "edge" layer in the power Internet of Things and uploads the operation log information to the edge disposal module;

[0023] An edge disposal module that filters out abnormal log information from the received operation log information, performs failure judgment on the abnormal log information to obtain the corresponding failure types, and executes the disposal strategies corresponding to the failure types. If the execution of the disposal strategy fails, this abnormal log information is sent to the Internet of Things management platform.

[0024] In a third aspect, the present invention also provides a multi-dimensional monitoring and handling method for an electric power Internet of Things based on cloud-edge collaboration, which is executed in an Internet of Things management platform and includes the following processes:

[0025] Receive the abnormal log information uploaded by the edge device;

[0026] Perform a fault judgment on the abnormal log information to obtain the corresponding fault type, and send the handling strategy corresponding to the fault type to the edge device for execution. If the execution of the handling strategy fails, record this abnormal log information in the fault information database.

[0027] Optionally, the performing a fault judgment on the abnormal log information includes:

[0028] Perform a fault judgment on the abnormal log information by using a fault judgment model trained based on the depthwise separable convolutional neural network algorithm.

[0029] Optionally, the training data of the fault judgment model is: a data set generated by using a GAN based on the abnormal log information in the fault information database.

[0030] Optionally, the judgment process for the failure of the execution of the handling strategy includes:

[0031] If the same abnormal log information is repeatedly received within a specified time after the execution of the handling strategy, it is determined that the execution of the handling strategy fails.

[0032] In a fourth aspect, the present invention also provides an Internet of Things management platform, including:

[0033] A cloud handling module, which receives the abnormal log information uploaded by the edge device; performs a fault judgment on the abnormal log information to obtain the corresponding fault type, and sends the handling strategy corresponding to the fault type to the edge device for execution. If the execution of the handling strategy fails, record this abnormal log information in the fault information database.

[0034] Optionally, it further includes a cloud management module, which is used to record and display the information on the successful execution of the handling strategy uploaded by the edge device, the information on the successful execution of the handling strategy of the cloud handling module, and the abnormal information that has not been handled, and add and update the abnormal log information, fault type, and handling strategy to the fault information database.

[0035] Compared with the prior art, the beneficial effects achieved by the present invention are:

[0036] 1) Fast handling speed. Since the edge device is closer to the service terminal, by performing abnormal analysis and handling at the edge end, the pressure of data explosion and network traffic can be reduced, and the operation and maintenance efficiency can be improved.

[0037] 2) Intelligent analysis and computing. Artificial intelligence algorithms are used to collaboratively process abnormal logs in the cloud and on the edge, providing the ability for intelligent data analysis and policy handling, and achieving edge area autonomy.

[0038] 3) Low cost. Data analysis and processing through edge computing eliminates the need to upload all data to the cloud, saving a lot of network and energy costs. Intelligent fault handling through cloud-edge collaboration also effectively reduces the manpower and material costs of on-site troubleshooting. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] Figure 1 is an architecture diagram of the system of the present invention;

[0040] Figure 2 The structure diagram of the deep separable convolutional neural network model: (a) is layer-by-layer convolution, (b) is point-by-point convolution;

[0041] Figure 3 Diagram of the training process of the fault analysis model based on the generative adversarial network. DETAILED DESCRIPTION

[0042] The present invention will be further described below in conjunction with the accompanying drawings. The following embodiments are only used to more clearly illustrate the technical solution of the present invention, and cannot be used to limit the protection scope of the present invention.

[0043] Example 1

[0044] The technical concept of the present invention is: based on multi-dimensional operation log information, the edge device close to the terminal device side remotely collaborates with the cloud management platform, and uses artificial intelligence algorithms to perform more accurate real-time analysis and disposal of network faults, provide intelligent edge autonomy capabilities, and be able to execute corresponding disposal strategies according to the fault type.

[0045] Intelligent monitoring and disposal with cloud-edge collaboration: Based on multi-dimensional monitoring log information, the power Internet of Things remotely collaborates with the cloud management platform with edge devices close to the terminal, delegates cloud computing tasks to edge devices, and uses edge computing capabilities to localize some abnormal logs and execute corresponding disposal strategies, thereby providing intelligent edge autonomy capabilities; at the same time, for a large amount of log information that is difficult to process, it can be uploaded to the cloud management platform, relying on platform resources for accurate intelligent fault analysis and disposal. The collaborative processing of network abnormalities by the platform and edge devices can effectively reduce data transmission pressure and reduce fault response time.

[0046] The system of the present invention is based on the power Internet of Things system with a three-layer architecture of "cloud-edge-end" in the prior art. Figure 1As shown in the figure, the "terminal" layer includes various sensors and acquisition terminals, which are used to collect on-site operation information of various power business scenarios such as power transmission, power distribution, and power consumption; the "edge" layer includes various containerized business APPs, edge devices, and edge management software. The business APPs collect the acquisition information of various sensors and acquisition terminals in the "terminal" layer and upload it to the edge management software, and the edge management software uploads the information to the "cloud" layer; the "cloud" layer includes an Internet of Things management platform, which analyzes the acquisition information and can feedback the analysis results to the "edge" layer.

[0047] A multi-dimensional monitoring and disposal system for power Internet of Things based on cloud-edge collaboration of the present invention is used to realize intelligent monitoring and fault disposal of edge devices in the power Internet of Things. Refer to Figure 1 As shown in the figure, it specifically includes:

[0048] 1) Edge monitoring module

[0049] The edge monitoring module is used to collect the operation log information of each edge device in the power Internet of Things, monitor the network operation status, and its collection frequency is at the minute level. The collected operation log information is shown in Table 1. The operation log information collected by this module includes the operation status information of edge devices, edge management software, containers, and business application APPs, which is obtained from the edge management software of the edge devices.

[0050] Among them, the operation information of edge devices specifically includes: CPU usage rate, memory usage rate, disk space usage rate, device operation duration, whether the device is abnormally restarted;

[0051] The operation information of the edge management software specifically includes: whether the software is abnormally restarted;

[0052] The operation information of containers specifically includes: CPU usage rate, memory usage rate, disk usage rate, whether the container is abnormally restarted;

[0053] The operation information of business application APPs specifically includes: CPU usage rate, memory usage rate, disk usage rate, whether the application is abnormally restarted, business data parsing situation, and terminal device offline situation.

[0054] This module calls the function functions provided by the edge management software and collects logs based on the MQTT message transmission protocol. Each type of log information corresponds to a message topic. Devices, containers, applications, etc. collected in the Internet of Things report log information to the corresponding topic at the same minute-level frequency. The log information includes various monitoring indicators, numbers, and attribute values of the collected objects. The edge monitoring module listens to all log information topics at all times. When all relevant log information in this time period is received and sorted in JSON format, the log information is sent to the edge disposal module.

[0055] Table 1 Collected operation logs

[0056]

[0057]

[0058] 2) Edge Disposal Module

[0059] The edge disposal module is used to analyze the fault type of the operation log information sent by the edge monitoring module through a fault judgment model in the edge device and execute the corresponding disposal strategy according to the fault type. This module utilizes the proximity analysis and computing capabilities provided by the edge device. Its "proximity" determines its ability to respond quickly, reduce the latency caused by network transmission, and achieve real-time computing.

[0060] In the edge disposal module, a fault judgment model (or random forest algorithm model) trained based on the random forest algorithm has been deployed. The training data of this model is the log information corresponding to high-frequency abnormal situations and the fault types corresponding to these abnormalities. The fault types are shown in Table 2, and there is a corresponding disposal strategy for each fault type. The disposal strategies include restarting the edge device, restarting the management software, restarting the container, shutting down the container, shutting down the end device, etc. In Table 2, the subscript i represents the i-th container, and the subscript ij represents the j-th application in the i-th container.

[0061] Table 2 Fault Types

[0062]

[0063]

[0064] The specific processing process of the edge disposal module is as follows: When receiving the operation log information sent by the edge monitoring module, it determines whether this log is an abnormal log according to the attribute information in the log (if the attribute value of a certain index exceeds the given normal threshold, it is judged as abnormal). The non-abnormal log information will be directly uploaded to the cloud management module in the IoT management platform for recording for subsequent query and management. The screened abnormal log data is then input into the fault judgment model (random forest algorithm model), and the model will output a fault type. Then, the corresponding disposal strategy is executed on the edge device side according to this fault type. If, after the execution of the disposal strategy, the edge disposal module repeatedly receives the same abnormal log information within the specified time, it is judged that the execution of this disposal strategy fails, and the abnormal information is sent to the cloud disposal module. Otherwise, the edge device reports the successful execution information of the disposal strategy to the cloud management module in the physical management platform, and the edge side always maintains cooperation with the cloud side.

[0065] 3) Cloud Disposal Module

[0066] The cloud disposal module is used to analyze the fault type according to the abnormal log information uploaded by the edge disposal module in the IoT management platform and issue the corresponding disposal strategy.

[0067] This module is used to analyze and solve abnormal situations that cannot be solved by the edge disposal module, and achieve efficient cooperation between the cloud and the edge. This module deploys a fault judgment model (or called a depthwise separable convolutional neural network model) trained by the depthwise separable convolutional neural network algorithm. This model replaces the standard convolutional layer of the traditional convolutional neural network with a depthwise separable convolutional layer, thereby greatly reducing network parameters, effectively reducing the training calculation amount and improving the algorithm efficiency. The depthwise separable convolutional layer consists of two parts: depthwise convolution and pointwise convolution. As Figure 2 shown, where (a) is depthwise convolution and (b) is pointwise convolution. First, for an input with dimensions of M×N×P 1 , a convolutional kernel of L×L×P 1 is used for depthwise convolution (generally, the value of L is usually taken as 3). The output feature map of this operation also has dimensions of M×N×P 1 ; then, pointwise convolution is used to fuse information from different channels. When the output dimension is M×N×P 2 , the number of parameters of pointwise convolution is only P 1 ×P 2 . The depthwise separable convolutional neural network model utilizes the sufficient computing resources and storage resources of the cloud platform, and based on a large number of abnormal logs and corresponding fault types as training data, after data preprocessing (including missing value processing, feature encoding, normalization, regularization, and feature selection, etc.), the fault judgment model is trained driven by data.

[0068] The specific processing process of the cloud disposal module is as follows: when the module receives the abnormal log information sent by the edge disposal module, it first performs data preprocessing on the abnormal log information (including missing value processing, feature encoding, normalization, regularization, and feature selection, etc.), and inputs the preprocessed abnormal log data into the fault judgment model (depthwise separable convolutional neural network model). The model will output the probability values corresponding to all fault types. The range of this probability value is from 0 to 1, and the sum of all output probability results is 1. If the maximum value among all fault type probability values is greater than a specific threshold (which can be set by oneself, for example, 80%), the disposal strategy corresponding to this fault type will be sent to the edge disposal module for execution. Otherwise, it indicates that the fault judgment model cannot determine the fault type corresponding to this abnormal log information, and this abnormal log will be reported to the cloud management module.

[0069] After the disposal strategy is sent, the cloud disposal module monitors the execution result of this strategy at all times. If the cloud disposal module repeatedly receives the same abnormal log information within the specified time, indicating that the execution of the disposal strategy fails, the failed operation and the abnormal log information will be reported to the cloud management module for manual disposal.

[0070] 4) Cloud management module

[0071] A cloud management module, which is used to record and display the operation result information of the edge disposal module, the operation result information of the cloud disposal module, and the unprocessed exception log information.

[0072] This module records and displays the disposal result information of the edge disposal module, the disposal result information of the cloud disposal module, and the unprocessed exception log information, which is convenient for operation and maintenance personnel to query and process. After analyzing and processing the unprocessed exception log information, the operation and maintenance personnel add this log information, the fault type, and the disposal strategy to the fault information database. This module is responsible for using the latest fault information database to update the fault judgment models in the edge disposal module and the cloud disposal module regularly.

[0073] Generally, there are not many faults in actual operation, and there are not enough marked cases for each specific fault. Therefore, the historical data obtained from the real operating system is not rich enough, and the effect of using it to build the analysis system is not ideal. Therefore, this module uses a generative adversarial network (GAN, generative adversarial network) to solve the problem of insufficient real data.

[0074] GAN includes two independent models, namely a generator and a discriminator. The generator receives a random variable that follows a certain distribution and is used to capture the distribution of real data. The discriminator outputs 1 (from real data) and 0 (from generated data) respectively to distinguish real samples and generated samples. During the training process, the generator and the discriminator are used to generate and classify samples respectively, and the sample performance is improved adversarially. GAN only needs to input some real data and noise that follows a certain rule. Through the game between the generator and the discriminator, when the discriminator tends to be stable, the generator can obtain the abnormal log status information that tends to the real data distribution.

[0075] As Figure 3 shown, the training process steps of the GAN-based fault judgment model are as follows:

[0076] Step 1: Collect the historical abnormal log data collected from the edge monitoring module. First, perform data preprocessing on the log data (including missing value processing, feature encoding, and normalization, etc.), and then input the processed real data set x into GAN.

[0077] Step 2: Initialize the discriminator parameter ω and the generator parameter θ, initialize the noise component z, let ε be a random number that follows a uniform distribution on [0,1], and input z into the generator G θ (z) to generate simulated data Let the sampled data The loss function of GAN can be defined as where, represents the data generated by the generator, It is sampled from a data set composed of real data and generated data. E[] represents the operation of taking the expectation, and D() represents the variance. is the penalty term, and λ is the penalty parameter.

[0078] Step 3: Use the Adam optimizer to train the discriminator parameter ω. The Adam optimizer dynamically adjusts the learning rate α of each parameter through the first-order moment estimation and second-order moment estimation of the gradient. m represents the iteration batch size, and β 1 and β 2 are the exponential decay rates of the first-order moment estimation and second-order moment estimation respectively. The learning rate of each iteration is within a fixed range, and relatively stable parameter updates can be achieved.

[0079] Step 4: Use the Adam optimizer to train the generator parameter θ until the generator parameter converges.

[0080] Step 5: Perform the same feature processing on the data generated by the GAN as on the real data (including dimensionality reduction, removing redundant data, etc.). 80% of the data is used for model training, and 20% of the data is used for model testing. Among them, the random forest algorithm model of the edge disposal module is trained using the relevant data of high-frequency faults, and the depthwise separable convolutional neural network model of the cloud disposal module is trained using all the data.

[0081] Finally, based on a small amount of labeled data sets, a large number of reliable labeled data sets are obtained for the training of the fault judgment models in the cloud disposal module and the edge disposal module, greatly improving the analysis accuracy of the models.

[0082] As a power Internet of Things monitoring and disposal system, the present invention realizes multi-dimensional monitoring and edge autonomy through the collection and analysis of various types of log information, and realizes accurate and intelligent analysis and rapid disposal of faults based on cloud-edge collaboration, effectively improving the operation and maintenance efficiency.

[0083] Embodiment 2

[0084] Based on the same inventive concept as in Embodiment 1, a multi-dimensional monitoring and disposal method for a power Internet of Things based on cloud-edge collaboration, which is executed in an edge device, includes the following processes:

[0085] Collect the operation log information of each edge device in the power Internet of Things;

[0086] Screen out the abnormal log information from the operation log information, perform fault judgment on the abnormal log information to obtain the corresponding fault type, and execute the disposal strategy corresponding to the fault type. If the execution of the disposal strategy fails, then send this abnormal log information to the IoT management platform.

[0087] Optionally, the failure judgment on the abnormal log information includes:

[0088] Using a failure judgment model trained based on the random forest algorithm to perform failure judgment on the abnormal log information.

[0089] Embodiment 3

[0090] Based on the same inventive concept as in Embodiment 2, an edge device of the present invention includes:

[0091] An edge monitoring module that collects the operation log information of each edge device in the power Internet of Things and uploads the operation log information to the edge disposal module;

[0092] An edge disposal module that filters out abnormal log information from the received operation log information, performs failure judgment on the abnormal log information to obtain the corresponding failure type, and executes the disposal strategy corresponding to the failure type. If the execution of the disposal strategy fails, this abnormal log information is sent to the Internet of Things management platform.

[0093] Embodiment 4

[0094] Based on the same inventive concept as in Embodiment 2, a multi-dimensional monitoring and disposal method for a power Internet of Things based on cloud-edge collaboration, which is executed in the Internet of Things management platform, includes the following processes:

[0095] Receiving the abnormal log information uploaded by the edge device;

[0096] Performing failure judgment on the abnormal log information to obtain the corresponding failure type, and sending the disposal strategy corresponding to the failure type to the edge device for execution. If the execution of the disposal strategy fails, this abnormal log information is recorded in the failure information database.

[0097] Optionally, the failure judgment on the abnormal log information includes:

[0098] Using a failure judgment model trained based on the depthwise separable convolutional neural network algorithm to perform failure judgment on the abnormal log information.

[0099] Embodiment 5

[0100] Based on the same inventive concept as in Embodiment 4, an Internet of Things management platform of the present invention includes:

[0101] A cloud disposal module that receives the abnormal log information uploaded by the edge device; performs failure judgment on the abnormal log information to obtain the corresponding failure type, and sends the disposal strategy corresponding to the failure type to the edge device for execution. If the execution of the disposal strategy fails, this abnormal log information is recorded in the failure information database.

[0102] Optionally, it further includes a cloud management module, which is used to record and display the successful information of the execution of the disposal strategy uploaded by the edge device, the successful information of the execution of the disposal strategy of the cloud disposal module, and the unhandled exception information, and add and update the exception log information, fault type, and disposal strategy to the fault information library.

[0103] Those skilled in the art should understand that the embodiments of the present application can be provided as a method, a system, or a computer program product. Therefore, the present application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present application can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0104] The present application is described with reference to the flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or block in the flowchart and / or block diagram can be implemented by computer program instructions, and the combination of the processes and / or blocks in the flowchart and / or block diagram can also be implemented by computer program instructions. These computer program instructions can be provided to the processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing devices to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing devices generate a device for implementing the specified function in one process Figure 1 one process or multiple processes and / or blocks Figure 1 or multiple blocks.

[0105] These computer program instructions can also be stored in a computer-readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer-readable memory generate a manufactured article including an instruction device, and the instruction device implements the specified function in one process Figure 1 one process or multiple processes and / or blocks Figure 1 or multiple blocks.

[0106] These computer program instructions can also be loaded onto a computer or other programmable data processing device, so that a series of operation steps are executed on the computer or other programmable device to generate a computer-implemented process. Therefore, the instructions executed on the computer or other programmable device provide steps for implementing the specified function in one process Figure 1 one process or multiple processes and / or blocks Figure 1 or multiple blocks.

[0107] The above are only the preferred embodiments of the present invention. It should be noted that for those of ordinary skill in the art, without departing from the technical principle of the present invention, several improvements and modifications can be made, and these improvements and modifications should also be regarded as the protection scope of the present invention.

Claims

1. A multi - dimensional monitoring and handling method for the power Internet of Things based on cloud - edge collaboration, which is executed in edge devices, characterized in that, it includes the following processes: Collect the operation log information of the "edge" layer in the power Internet of Things; among them, the operation log information includes the operation status information of edge devices, edge management software, containers, and business application APPs, where: The operation information of edge devices specifically includes: CPU usage rate, memory usage rate, disk space usage rate, device operation duration, and whether the device has abnormal restarts; The operation information of edge management software specifically includes: whether the software has abnormal restarts; The operation information of containers specifically includes: CPU usage rate, memory usage rate, disk usage rate, and whether the container has abnormal restarts; The operation information of business application APPs specifically includes: CPU usage rate, memory usage rate, disk usage rate, whether the application has abnormal restarts, business data parsing situation, and end - device offline situation; Filter out abnormal log information from the operation log information, perform fault judgment on the abnormal log information to obtain the corresponding fault type, and execute the disposal strategy corresponding to the fault type. If the execution of the disposal strategy fails, send this abnormal log information to the Internet of Things management platform; Among them, the fault judgment on the abnormal log information includes: Using a fault judgment model trained based on the random forest algorithm to perform fault judgment on the abnormal log information; The training data of the fault judgment model is the log information corresponding to high - frequency abnormal situations and the fault types corresponding to the abnormalities.

2. A multi - dimensional monitoring and handling method for the power Internet of Things based on cloud - edge collaboration according to claim 1, characterized in that, The edge device includes: An edge monitoring module, which collects the operation log information of the "edge" layer in the power Internet of Things and uploads the operation log information to the edge disposal module; An edge disposal module, which filters out abnormal log information from the received operation log information, performs fault judgment on the abnormal log information to obtain the corresponding fault type, and executes the disposal strategy corresponding to the fault type. If the execution of the disposal strategy fails, send this abnormal log information to the Internet of Things management platform.

3. A multi - dimensional monitoring and handling method for the power Internet of Things based on cloud - edge collaboration, which is executed in the Internet of Things management platform, characterized in that, it includes the following processes: Receive the abnormal log information uploaded by the edge device; Perform fault judgment on the abnormal log information to obtain the corresponding fault type, and send down the disposal strategy corresponding to the fault type to the edge device for execution. If the execution of the disposal strategy fails, record this abnormal log information in the fault information database; Among them, the fault judgment on the abnormal log information includes: Using a fault judgment model trained based on the depth - separable convolutional neural network algorithm to perform fault judgment on the abnormal log information; the depth - separable convolutional layer consists of two parts: layer - by - layer convolution and point - wise convolution. First, for an input with a dimension of M×N×P1, use a convolution kernel of L×L×P1 for layer - by - layer convolution to obtain an output with a dimension of M×N×P1; then, fuse the information of different channels through point - wise convolution. When the output dimension is M×N×P2, the number of parameters of point - wise convolution is only P1×P2; Among them, the training data of the fault judgment model is: a data set generated by using GAN based on the abnormal log information in the fault information library; Among them, GAN includes two independent models, a generator and a discriminator; the generator receives a random variable subject to a certain distribution and is used to capture the distribution of real data; the discriminator outputs 1 and 0 respectively to distinguish real samples and generated samples; The training process steps of the fault judgment model based on GAN are as follows: Step 1: Collect the historical abnormal log data collected from the edge monitoring module. First, perform data preprocessing on the log data, and input the processed real data set x into GAN; Step 2: Initialize the discriminator parameters ω and the generator parameters θ, initialize the noise component z, let ε be a random number uniformly distributed in [0, 1], and input z into the generator G θ (z) to generate simulated data Let the sampled data The loss function of GAN is defined as where represents the data generated by the generator, is sampled from the dataset composed of real data and generated data, E[] represents the operation of taking the expectation, D() represents the variance, is the penalty term, and λ is the penalty parameter; Step 3: Use the Adam optimizer Train the discriminator parameters ω. The Adam optimizer dynamically adjusts the learning rate α of each parameter through the first-order moment estimate and second-order moment estimate of the gradient. m represents the iteration batch size, and β 1 , β 2 are the exponential decay rates of the first-order moment estimate and second-order moment estimate respectively; Step 4: Use the Adam optimizer Train the generator parameters θ until the generator parameters converge; Step 5: Perform the same feature processing on the data generated by GAN as on the real data. 80% of the data is used for model training, and 20% of the data is used for model testing. Among them, the relevant data of high-frequency faults is used to train the random forest algorithm model of the edge disposal module, and all the data is used to train the depthwise separable convolutional neural network model of the cloud disposal module.

4. A multi-dimensional monitoring and disposal method for a power Internet of Things based on cloud-edge collaboration according to claim 3, characterized in that The Internet of Things management platform includes: A cloud disposal module, which receives the abnormal log information uploaded by the edge device; performs fault judgment on the abnormal log information to obtain the corresponding fault type, and issues the disposal strategy corresponding to the fault type to the edge device for execution. If the execution of the disposal strategy fails, record this abnormal log information in the fault information library.

5. A multi-dimensional monitoring and disposal method for a power Internet of Things based on cloud-edge collaboration according to claim 4, characterized in that The Internet of Things management platform further includes: A cloud management module, which is used to record and display the information of successful execution of the disposal strategy and the abnormal information that has not been disposed of, and add and update the abnormal log information, fault type and disposal strategy to the fault information library.

Images