Adaptive system and method for end-side cloud dynamic task scheduling of power distribution Internet of Things

Abstract

The invention discloses a self-adaptive system and method for end-side cloud dynamic task scheduling of a power distribution internet of things, and relates to the technical field of cloud computing regulation and control, and the method comprises the steps: collecting power distribution network environment data and worker operation data through a power distribution network equipment sensor connected through a gateway; carrying out preprocessing and feature extraction on the collected data to generate a standardized data stream; dynamically distributing calculation tasks to a local gateway or a cloud server according to a worker request type and an equipment load state; dynamically adjusting a resource allocation strategy based on the network bandwidth of the power distribution network, the computing power of the equipment and the task priority; monitoring data transmission safety and equipment state abnormity of the cloud computing environment of the power distribution network in real time, and triggering a protection mechanism; and displaying the data processing result, the task state and the security alarm information through a gateway interface. The method has the advantages that by dynamically distributing local and cloud computing resources, the task processing efficiency and the system adaptability under the power distribution network multi-device cooperation scene are remarkably improved.

Description

Technical Field

[0001] The present invention relates to the technical field of cloud computing regulation and control, and specifically relates to an adaptive system and method for dynamic task scheduling of the edge-cloud in a distribution Internet of Things. Background Art

[0002] With the popularization of the Internet of Things, the amount of sensor data in the distribution network scenario has increased sharply and the device heterogeneity is significant. Traditional solutions mostly rely on a single cloud processing mode, resulting in limited real-time response capabilities. In the prior art, the distribution network gateway device only undertakes the data transfer function and fails to effectively coordinate local computing resources and cloud computing power. Especially in scenarios such as media rendering and environmental analysis, problems such as network bandwidth congestion and device load imbalance are likely to occur. In addition, the existing system lacks a dynamic task migration mechanism for the distribution network scenario and is difficult to adapt to complex working conditions such as device offline and sudden changes in computing requirements in a timely manner, resulting in double losses of staff experience and energy efficiency. Summary of the Invention

[0003] To solve the above technical problems, an adaptive system and method for dynamic task scheduling of the edge-cloud in a distribution Internet of Things are provided. The technical solution solves the problem that the existing system lacks a dynamic task migration mechanism for the distribution network scenario and is difficult to adapt to complex working conditions such as device offline and sudden changes in computing requirements in a timely manner, resulting in double losses of experience and energy efficiency.

[0004] To achieve the above object, the technical solution adopted by the present invention is as follows:

[0005] An adaptive method for dynamic task scheduling of the edge-cloud in a distribution Internet of Things, comprising:

[0006] Collecting distribution network environment data and staff operation data through sensors of distribution network devices connected by a gateway;

[0007] Preprocessing and feature extraction of the collected data to generate a standardized data stream;

[0008] Dynamically allocating computing tasks to the local gateway or the cloud server according to the staff request type and the device load status;

[0009] Based on the distribution network bandwidth, device computing power and task priority, dynamically adjusting the resource allocation strategy;

[0010] Real-time monitoring of the data transmission security and device status anomalies in the distribution network cloud computing environment and triggering a protection mechanism;

[0011] Displaying the data processing results, task status and security warning information through the gateway interface.

[0012] Preferably, the collecting distribution network environment data and staff operation data through sensors of distribution network devices connected by a gateway specifically includes:

[0013] Collect the environmental parameters of the distribution network through temperature and light sensors, and collect application operation logs through the staff terminal device;

[0014] Build an encryption module in the gateway to add dynamic tags and access permission identifiers to the collected sensitive data;

[0015] Set the data collection timing window and update the data set according to the preset period or event trigger mode.

[0016] Preferably, the preprocessing and feature extraction of the collected data to generate a standardized data stream specifically includes:

[0017] Perform noise reduction and normalization processing on the original data to generate a multi-dimensional feature vector;

[0018] Use a convolutional neural network to perform real-time classification on the media stream data to identify staff preferences and abnormal operation patterns;

[0019] Package the processed data into a structured data packet and attach a timestamp and device identifier.

[0020] Preferably, the dynamic allocation of computing tasks to the local gateway or cloud server according to the staff request type and device load status specifically includes:

[0021] Calculate the device-task allocation priority according to the task type and device computing power. Media rendering tasks are preferentially allocated to local GPU resources, and batch data analysis tasks are preferentially uploaded to the cloud;

[0022] When it is detected that the gateway load exceeds the threshold, start the load transfer algorithm and migrate some tasks to other intelligent devices within the distribution network local area network;

[0023] Adopt a dynamic queue management mechanism to update the scheduling strategy in real time according to the task completion status.

[0024] Preferably, the load transfer algorithm adopts a device reputation scoring mechanism and preferentially allocates load tasks to high-reputation devices. Specifically, the calculation method of the device reputation score is:

[0025]

[0026] Where, is the device reputation score of the i-th device, is the average historical task completion rate of the i-th device, is the average communication delay between the i-th device and the gateway, is the maximum tolerable value of the communication delay, is the weight coefficient.

[0027] Preferably, the dynamic adjustment of the resource allocation strategy based on the network bandwidth, device computing power, and task priority of the distribution network specifically includes:

[0028] Based on the computing tasks to be processed, combined with the computing power of all intelligent devices within the distribution network local area network, generate several computing resource allocation plans;

[0029] Using the network bandwidth occupancy rate and device-task allocation priority as evaluation factors, based on the TOPSIS algorithm, select the optimal computing resource allocation plan from all computing resource allocation plans;

[0030] Adopt a reinforcement learning algorithm to iteratively optimize the resource allocation plan, and adjust the device-task allocation priority according to the feedback of historical task execution efficiency.

[0031] Preferably, the real-time monitoring of the data transmission security and device status anomalies in the distribution network cloud computing environment and triggering the protection mechanism specifically includes:

[0032] Embed a two-way authentication protocol in the data transmission link to throttle or block abnormal access requests;

[0033] Establish a device behavior baseline through the LSTM network. When the deviation of the operation sequence is detected to exceed the threshold, generate a risk warning and isolate the suspicious device;

[0034] Regularly push security reports to the staff terminal.

[0035] Preferably, the display of the data processing results, task status, and security warning information through the gateway interface specifically includes:

[0036] Render a 3D distribution network environment map in the gateway interface, and dynamically display the data streams, task execution progress, and security status of each device;

[0037] Support voice and gesture interaction, and switch the data view or adjust the resource allocation plan according to the staff's instructions;

[0038] When a high-risk warning is triggered, automatically pop up an emergency handling interface and provide a multi-factor authentication channel.

[0039] Furthermore, an adaptive system for dynamic task scheduling at the edge and cloud of the distribution Internet of Things is proposed, which is used to implement the adaptive method for dynamic task scheduling at the edge and cloud of the distribution Internet of Things as described above, including:

[0040] A data acquisition module, which is used to collect the distribution network environment parameters and operation logs through temperature sensors, light sensors, and staff terminal devices, and add dynamic tags and access permission identifiers to sensitive data in the built-in encryption module;

[0041] A preprocessing module for performing noise reduction, normalization, and real-time classification based on a convolutional neural network on the original data to generate a structured data packet with a timestamp;

[0042] A dynamic task allocation module for allocating computing tasks to the local GPU, cloud server, or other intelligent devices within the distribution network local area network according to the device load threshold and task type through a load transfer algorithm;

[0043] A resource scheduling optimization module for iteratively generating an optimal allocation scheme for network bandwidth and device priority based on the TOPSIS algorithm and reinforcement learning;

[0044] A security protection module for implementing abnormal traffic blocking and device isolation through a two-way authentication protocol and LSTM network behavior baseline detection;

[0045] An interactive interface module for rendering a 3D distribution network environment map and supporting voice / gesture operations, and a real-time display of the task progress and an emergency verification interface for high-risk warnings.

[0046] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0047] The adaptive method for dynamic task scheduling of the distribution Internet of Things edge-cloud provided by the present invention significantly improves the task processing efficiency and system adaptability in the multi-device collaboration scenario of the distribution network by dynamically allocating local and cloud computing resources. On the one hand, the intelligent scheduling mechanism based on the device load status and task type can flexibly call the local GPU computing power and cloud distributed resources, reducing the processing delay of media rendering tasks while optimizing the network bandwidth utilization rate, and effectively alleviating the response lag problem caused by single reliance on the cloud in the traditional scheme. On the other hand, combined with the device behavior pattern analysis and dynamic security authentication mechanism, abnormal operations and potential attack behaviors can be accurately identified, and risk nodes can be quickly isolated through multi-factor verification, greatly enhancing the active protection ability of the privacy data of the distribution network. In addition, through the three-dimensional visualization interface and multi-modal interaction design, intuitive device status monitoring and resource scheduling decision support are provided for the staff, realizing a highly robust cloud computing service experience for the distribution network while reducing the operation complexity. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] Figure 1 It is a flowchart of the adaptive method for dynamic task scheduling of the distribution Internet of Things edge-cloud proposed by the present invention;

[0049] Figure 2 It is a flowchart of the method for collecting distribution network environment data and staff operation data proposed by the present invention;

[0050] Figure 3 It is a flowchart of the method for generating a standardized data stream proposed by the present invention;

[0051] Figure 4 Flow chart of the method for dynamically allocating computing tasks proposed by the present invention;

[0052] Figure 5 Flow chart of the method for dynamically adjusting resource allocation strategies proposed by the present invention;

[0053] Figure 6 Flow chart of the method for real-time monitoring of data transmission security and equipment status anomalies in the cloud computing environment of the distribution network proposed by the present invention;

[0054] Figure 7 Flow chart of the method for displaying data processing results, task status, and security warning information through the gateway interface proposed by the present invention. Specific embodiments

[0055] The following description is used to disclose the present invention so that those skilled in the art can implement the present invention. The preferred embodiments in the following description are only examples, and those skilled in the art can think of other obvious variations.

[0056] Refer to Figure 1 As shown, an adaptive method for dynamic task scheduling of the distribution Internet of Things edge-cloud includes:

[0057] Collect distribution network environment data and staff operation data through distribution network device sensors connected by the gateway;

[0058] By integrating multi-source sensors and staff terminal devices, full-dimensional collection of distribution network environment parameters and staff behaviors is achieved, avoiding the problem of data islands in traditional solutions;

[0059] Preprocess and extract features from the collected data to generate a standardized data stream;

[0060] Through noise reduction, normalization, and feature extraction based on deep learning, redundancy and noise interference in the original data are eliminated, improving data quality; Structured data packet encapsulation combined with timestamps and device identifiers ensures the spatio-temporal consistency of multi-source heterogeneous data, significantly improving the accuracy of subsequent task allocation and resource scheduling;

[0061] Dynamically allocate computing tasks to the local gateway or cloud server according to the staff request type and device load status;

[0062] Intelligent scheduling based on task type and real-time load gives full play to the real-time rendering advantage of the local GPU and the large-scale computing power of the cloud, taking into account the low-latency and high-throughput requirements; Through the load transfer algorithm and dynamic queue management, system jams caused by overloading of a single node are avoided, and service stability in complex scenarios is improved;

[0063] Dynamically adjust the resource allocation strategy based on the network bandwidth, device computing power, and task priorities of the distribution network;

[0064] Through multi-dimensional evaluation that integrates network status, device computing power, and task urgency, achieve global optimization of resource allocation to ensure instant response to high-priority tasks such as security alarms; combine dynamic policy iteration of reinforcement learning to adapt to the addition or deletion of distribution network devices or network fluctuations, significantly enhancing the system's resilience and long-term operation efficiency;

[0065] Real-time monitor the data transmission security and device status anomalies in the distribution network cloud computing environment, and trigger the protection mechanism;

[0066] Based on behavior baseline modeling and two-way authentication protocols, actively identify threats such as DDoS attacks and data tampering, and achieve millisecond-level anomaly blocking and device isolation; through the security alarm and emergency verification linkage mechanism, reduce the impact scope of malicious intrusion on the distribution network, and ensure the continuous availability of critical services;

[0067] Display the data processing results, task status, and security alarm information through the gateway interface.

[0068] With a visual map and multi-modal interaction design, intuitively present the device operation status and data flow topology, reducing the cognitive burden on staff; combined with the automatic pop-up window and voice prompt of security alarms, achieve hierarchical push of risk information, enhance the staff's real-time perception and proactive handling ability of system anomalies, and improve the overall human-machine collaboration efficiency.

[0069] Refer to Figure 2 As shown, the distribution network device sensors connected through the gateway collect distribution network environment data and staff operation data, specifically including:

[0070] Collect distribution network environment parameters through temperature and light sensors, and collect application operation logs through staff terminal devices;

[0071] Install an encryption module in the gateway to add dynamic tags and access permission identifiers to the collected sensitive data;

[0072] Set the data collection time sequence window, and update the data set according to the preset cycle or event trigger mode.

[0073] By collaboratively collecting environmental and behavioral data through temperature and light sensors and staff terminal devices, multi-dimensional perception of physical environmental parameters and staff digital operation trajectories is achieved, breaking through the limitations of traditional single data sources and providing holographic data support for distribution network scenario analysis. By building an encryption module into the gateway and dynamically marking permissions for sensitive data, a differentiated security management and control mechanism for the data collection stage is constructed, which can not only ensure the compliance flow of staff privacy data, but also avoid resource waste caused by global encryption. Combined with the event triggering and periodic update strategy of the timing window, it effectively balances the conflict between real-time data collection and system resource consumption, ensuring immediate response to emergency scenarios (such as abnormal high temperature alarms) and low-power operation of normal monitoring, and realizing the coordinated optimization of data collection efficiency and equipment endurance.

[0074] Reference Figure 3 As shown, the preprocessing and feature extraction of the collected data to generate a standardized data stream specifically includes:

[0075] Perform noise reduction and normalization processing on the original data to generate a multi-dimensional feature vector;

[0076] Use convolutional neural networks to classify media stream data in real time to identify staff preferences and abnormal operating patterns;

[0077] The processed data is encapsulated into a structured data packet with a timestamp and device identifier.

[0078] Through noise reduction and normalization processing, the data deviation caused by sensor noise and device heterogeneity is effectively eliminated, significantly improving the accuracy of subsequent task analysis; using convolutional neural networks to perform real-time feature extraction and pattern recognition on media streams, it not only realizes the dynamic update of staff preference portraits to support personalized service recommendations, but also simultaneously completes millisecond-level detection of abnormal operation behaviors, strengthening the timeliness of distribution network data security protection. Encapsulating multi-source data into structured data packets with time and space identifiers not only solves the problem of device data format fragmentation, but also provides accurate contextual association information for resource scheduling algorithms through the association mapping of timestamps and device identifiers, so that the dynamic adaptation efficiency of cloud collaborative computing task allocation and network bandwidth is systematically optimized.

[0079] Reference Figure 4 As shown, the dynamic allocation of computing tasks to the local gateway or cloud server according to the staff request type and the device load status specifically includes:

[0080] Calculate the device-task allocation priority according to the task type and device computing power. Among them, the device-task allocation priority of batch data analysis tasks and cloud devices is 1, and the device-task allocation priority of media rendering tasks and local GPU devices is higher than that of media rendering tasks and cloud devices, so as to realize that media rendering tasks are preferentially allocated to local GPU resources and batch data analysis tasks are preferentially uploaded to the cloud. Specifically, the task type and device computing power are used to calculate the device-task allocation priority, which is allocated through task execution experience. For example, devices with high-performance GPUs / TPUs are suitable for compute-intensive tasks, and devices with large-capacity hard disks are suitable for data analysis tasks;

[0081] When it is detected that the gateway load exceeds the threshold, start the load transfer algorithm and migrate some tasks to other intelligent devices within the distribution network local area network;

[0082] Adopt a dynamic queue management mechanism and update the scheduling policy in real time according to the task completion status.

[0083] The load transfer algorithm adopts a device reputation scoring mechanism and preferentially allocates load tasks to high-reputation devices. Specifically, the calculation method of device reputation scoring is as follows:

[0084]

[0085] In the formula, is the device reputation score of the i-th device, is the average historical task completion rate of the i-th device, is the average communication delay between the i-th device and the gateway, is the maximum tolerable value of communication delay, is the weight coefficient.

[0086] The dynamic task allocation mechanism significantly improves the resource utilization efficiency of the distribution network cloud computing system through the precise matching of task types and device capabilities. Based on the device-task allocation priority strategy, compute-intensive tasks such as media rendering are preferentially scheduled to local GPU / TPU devices, which can give full play to the hardware acceleration ability to reduce rendering latency; while batch data analysis tasks are directed to the cloud or large-capacity storage devices, effectively avoiding processing blockages caused by insufficient local computing power. When the gateway load exceeds the limit, the load transfer algorithm combines the device reputation scoring mechanism and preferentially selects intelligent devices with high task completion rates and low communication delays as migration nodes, which not only ensures the reliability of task execution but also avoids the risk of secondary latency caused by network fluctuations. Through the collaborative optimization of dynamic queue management and reputation scoring, the system can adapt to the online state changes of distribution network devices and continuously maintain the optimal state of task scheduling and load balancing in the heterogeneous resource pool, significantly enhancing the service robustness in complex distribution network scenarios.

[0087] Refer to Figure 5 As shown, the specific steps of dynamically adjusting the resource allocation strategy based on the network bandwidth, device computing power and task priority of the distribution network are as follows:

[0088] Based on the computing tasks to be processed, combined with the computing power of all intelligent devices in the distribution network local area network, generate several computing resource allocation plans;

[0089] Using the network bandwidth occupancy rate and device-task allocation priority as evaluation factors, based on the TOPSIS algorithm, select the optimal computing resource allocation plan from all computing resource allocation plans;

[0090] Adopt the reinforcement learning algorithm to iteratively optimize the resource allocation plan, and adjust the device-task allocation priority according to the feedback of the historical task execution efficiency.

[0091] Specifically, the calculation formula for the evaluation coefficient of the network bandwidth occupancy rate is:

[0092]

[0093] Wherein, is the network bandwidth occupancy rate, is the estimated time for task completion, is the network bandwidth occupancy in the t-th period within the estimated time for task completion, is the total network bandwidth of the distribution network, is the damping coefficient;

[0094] The device-task allocation priority is obtained by accumulating the devices assigned to each task in the computing resource allocation plan and the device-task allocation priority of the task type, as the device-task allocation priority evaluation coefficient;

[0095] Perform positive normalization processing on the evaluation coefficient of the network bandwidth occupancy rate and the evaluation coefficient of the device-task allocation priority respectively, to obtain the standard evaluation coefficient of the network bandwidth occupancy rate and the standard evaluation coefficient of the device-task allocation priority. The larger the standard evaluation coefficient of the network bandwidth occupancy rate, the lower the network bandwidth occupancy rate within the estimated time for task completion. The larger the standard evaluation coefficient of the device-task allocation priority, the higher the matching degree between the task and the device assigned to it;

[0096] Select the maximum values of the standard evaluation coefficient of the network bandwidth occupancy rate and the standard evaluation coefficient of the device-task allocation priority as the global ideal optimal vector;

[0097] Select the minimum values of the standard evaluation coefficient of the network bandwidth occupancy rate and the standard evaluation coefficient of the device-task allocation priority as the global ideal worst vector;

[0098] Set the network bandwidth occupancy rate weight and the task processing efficiency weight respectively;

[0099] Combine the network bandwidth occupancy rate weight and the task processing efficiency weight, and calculate the TOPSIS evaluation value of each resource allocation scheme through the weighted TOPSIS formula;

[0100] The weighted TOPSIS formula is specifically as follows:

[0101]

[0102] In the formula, is the TOPSIS evaluation value of the jth resource allocation scheme, is the standard evaluation coefficient of the network bandwidth occupancy rate of the jth resource allocation scheme, is the standard evaluation coefficient of the device - task allocation priority of the jth resource allocation scheme, is the minimum value of the standard evaluation coefficient of the network bandwidth occupancy rate, is the minimum value of the standard evaluation coefficient of the device - task allocation priority, is the maximum value of the standard evaluation coefficient of the network bandwidth occupancy rate, is the maximum value of the standard evaluation coefficient of the device - task allocation priority, is the network bandwidth occupancy rate weight, is the task processing efficiency weight. The network bandwidth occupancy rate weight and the task processing efficiency weight are dynamically adjusted based on human attributes. When the task has a high requirement for processing timeliness, the task processing efficiency weight is increased to improve the matching degree between the task and the device, thereby improving the task processing efficiency. When the task is processed, if it is necessary to reduce the impact on the network usage of other network devices in the distribution network, the network bandwidth occupancy rate weight needs to be increased;

[0103] Select the resource allocation scheme corresponding to the maximum TOPSIS evaluation value as the optimal computing resource allocation scheme.

[0104] The dynamic resource allocation strategy significantly improves the intelligent level of resource scheduling in the heterogeneous computing environment of the distribution network through a multi - dimensional decision - making model. Based on the collaborative optimization of the network bandwidth occupancy rate and the device - task matching degree by the TOPSIS algorithm, it can quickly screen out the global optimal solution with the smallest bandwidth pressure and the highest task execution efficiency among the generated multiple sets of resource allocation schemes, effectively balancing the conflict between network transmission load and computing performance.

[0105] Refer to Figure 6 As shown, the real - time monitoring of the data transmission security and device status anomalies in the distribution network cloud computing environment and triggering the protection mechanism specifically include:

[0106] Implement a two-way authentication protocol in the data transmission link to impose traffic rate limits or block abnormal access requests;

[0107] Establish a device behavior baseline through an LSTM network. When the deviation of the operation sequence is detected to exceed the threshold, generate a risk alert and isolate the suspicious device;

[0108] Regularly push security reports to the staff terminals.

[0109] Through multi-level technology integration, significantly enhance the active defense ability of the distribution network cloud computing system. The two-way authentication protocol constructs a two-way identity verification mechanism in the data transmission link, effectively intercepting the access of forged devices and man-in-the-middle attacks. At the same time, combined with the intelligent rate limit strategy for abnormal traffic, while blocking malicious access, it maximally guarantees the continuity of normal services.

[0110] Based on the device behavior baseline modeling of the LSTM network, use its ability to capture temporal dependencies to identify hidden threats such as distributed denial-of-service attacks (DDoS) and zero-day vulnerability exploitation. Compared with traditional rule-based detection methods, it can discover more complex attack patterns and latent intrusion behaviors.

[0111] The immediate linkage between the suspicious device isolation mechanism and security alerts can quickly contain the lateral spread of attacks within the distribution network local area network and prevent critical devices from being infiltrated and manipulated. Combined with the regularly pushed customized security reports, the staff can clearly master the network risk situation and protection and disposal records of the distribution network, forming a security closed-loop of "monitoring - blocking - feedback", greatly reducing the occurrence probability of privacy data leakage and abnormal device downtime in the distribution network, and at the same time reducing the intervention cost of the staff's daily security operation and maintenance.

[0112] Refer to Figure 7 As shown, the display of data processing results, task status, and security alert information through the gateway interface specifically includes:

[0113] Render a 3D distribution network environment map in the gateway interface, and dynamically display the data streams, task execution progress, and security status of each device;

[0114] Support voice and gesture interactions, and switch data views or adjust resource allocation plans according to the staff's instructions;

[0115] When a high-risk alert is triggered, automatically pop up an emergency handling interface and provide a multi-factor authentication channel.

[0116] Through the multi-modal presentation and intelligent response mechanism, it significantly improves the overall control ability and emergency response efficiency of distribution network staff for cloud computing services. The dynamic rendering technology of the 3D distribution network environment map spatially visualizes the device data stream, task execution link, and security status, helping staff intuitively identify device operation bottlenecks or abnormal hot spots, and reducing the information understanding threshold in the scenario of multi-device collaboration. Voice and gesture interactions support non-contact operations, facilitating staff to quickly switch data views or adjust resource allocation strategies, especially suitable for the barrier-free use requirements in the distribution network environment. When a high-risk alarm is triggered, the emergency processing interface automatically pops up and embeds a multi-factor authentication channel, which not only ensures strict control of critical operation permissions but also avoids the distraction of the traditional verification process to the staff's attention, achieving a dual optimization of security protection and operation fluency. The overall design forms a closed-loop management of "status perception - decision intervention - risk disposal" through the reconstruction of the human-computer interaction logic, effectively enhancing the staff's security trust and active participation in the distribution network cloud service.

[0117] Furthermore, based on the same inventive concept as the above-mentioned adaptive method for dynamic task scheduling of the distribution Internet of Things edge-cloud, this solution proposes an adaptive system for dynamic task scheduling of the distribution Internet of Things edge-cloud, including:

[0118] A data acquisition module, which is used to collect distribution network environment parameters and operation logs through temperature sensors, light sensors, and staff terminal devices, and add dynamic tags and access permission identifiers to sensitive data in the built-in encryption module;

[0119] A preprocessing module, which is used to perform noise reduction, normalization, and real-time classification based on a convolutional neural network on the original data to generate a structured data packet with a timestamp;

[0120] A dynamic task allocation module, which is used to allocate computing tasks to the local GPU, cloud server, or other intelligent devices within the distribution network local area network through a load transfer algorithm according to the device load threshold and task type;

[0121] A resource scheduling optimization module, which is used to iteratively generate the optimal allocation scheme of network bandwidth and device priority based on the TOPSIS algorithm and reinforcement learning;

[0122] A security protection module, which is used to implement abnormal traffic blocking and device isolation through a two-way authentication protocol and an LSTM network behavior baseline detection;

[0123] An interaction interface module, which is used to render the 3D distribution network environment map and support voice / gesture operations, and display the task progress and the emergency verification interface for high-risk alarms in real time.

[0124] In summary, the advantages of the present invention are as follows: By dynamically allocating local and cloud computing resources, the task processing efficiency and system adaptability in the scenario of multi-device collaboration in the distribution network are significantly improved. On the one hand, based on the intelligent scheduling mechanism of device load status and task type, it can flexibly call local GPU computing power and cloud distributed resources, reduce the processing latency of media rendering tasks, optimize the network bandwidth utilization rate, and effectively alleviate the response lag problem caused by single reliance on the cloud in traditional solutions. On the other hand, combined with device behavior pattern analysis and dynamic security authentication mechanism, it can accurately identify abnormal operations and potential attack behaviors, and quickly isolate risk nodes through multi-factor verification, greatly enhancing the active protection ability of privacy data in the distribution network. In addition, through the three-dimensional visualization interface and multi-modal interaction design, it provides intuitive device status monitoring and resource scheduling decision support for staff, realizes a highly robust cloud computing service experience for the distribution network while reducing the operation complexity.

[0125] The foregoing has shown and described the basic principles, main features and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited by the above embodiments, and what is described in the above embodiments and the specification is only the principle of the present invention. Without departing from the spirit and scope of the present invention, the present invention will have various changes and improvements, and these changes and improvements all fall within the scope of the present invention claimed. The scope of protection required by the present invention is defined by the appended claims and their equivalents.

Claims

1. An adaptive method for dynamic task scheduling of distribution Internet of Things edge-cloud, characterized in that Including: Collecting distribution network environment data and staff operation data through sensors of distribution network equipment connected by the gateway; Preprocessing and feature extraction of the collected data to generate a standardized data stream; Dynamically allocating computing tasks to the local gateway or cloud server according to the staff request type and equipment load status; Dynamically adjusting the resource allocation strategy based on the distribution network bandwidth, equipment computing power, and task priority; Real-time monitoring of data transmission security and equipment status anomalies in the distribution network cloud computing environment and triggering a protection mechanism; Displaying data processing results, task status, and security warning information through the gateway interface.

2. An adaptive method for dynamic task scheduling of distribution Internet of Things edge-cloud, according to claim 1, characterized in that, The collection of distribution network environment data and staff operation data through sensors of distribution network equipment connected by the gateway specifically includes: Collecting distribution network environment parameters through temperature and light sensors, and collecting application operation logs through staff terminal devices; Installing an encryption module in the gateway to add dynamic tags and access permission identifiers to the collected sensitive data; Setting a data collection timing window and updating the data set according to a preset period or event trigger mode.

3. An adaptive method for dynamic task scheduling of the distribution Internet of Things edge-cloud, according to claim 2, characterized in that The preprocessing and feature extraction of the collected data to generate a standardized data stream specifically includes: Performing noise reduction and normalization processing on the original data to generate a multi-dimensional feature vector; Using a convolutional neural network to perform real-time classification on media stream data to identify staff preferences and abnormal operation patterns; Encapsulating the processed data into a structured data packet and attaching a timestamp and device identifier.

4. An adaptive method for dynamic task scheduling of distribution Internet of Things edge-cloud, according to claim 3, characterized in that The dynamic allocation of computing tasks to the local gateway or cloud server according to the staff request type and equipment load status specifically includes: Calculating the device-task allocation priority according to the task type and equipment computing power. Media rendering tasks are preferentially allocated to local GPU resources, and batch data analysis tasks are preferentially uploaded to the cloud; When it is detected that the gateway load exceeds the threshold, start the load transfer algorithm and migrate some tasks to other intelligent devices within the distribution network local area network; Adopting a dynamic queue management mechanism to update the scheduling strategy in real time according to the task completion status.

5. An adaptive method for dynamic task scheduling of the edge-cloud in a distribution Internet of Things according to claim 4, characterized in that, The load transfer algorithm adopts a device reputation scoring mechanism and preferentially allocates load tasks to high-reputation devices. Specifically, the calculation method of the device reputation score is: ； Wherein, is the device reputation score of the i-th device, is the average historical task completion rate of the i-th device, is the average communication delay between the i-th device and the gateway, is the maximum tolerable value of the communication delay, is the weight coefficient.

6. An adaptive method for dynamic task scheduling of distribution Internet of Things edge-cloud, according to claim 5, characterized in that, The dynamic adjustment of the resource allocation strategy based on the distribution network bandwidth, equipment computing power, and task priority specifically includes: Based on the computing tasks to be processed, combined with the computing power of all intelligent devices within the distribution network local area network, generating several computing resource allocation plans; Using the network bandwidth occupancy rate and device-task allocation priority as evaluation factors, and based on the TOPSIS algorithm, selecting the optimal computing resource allocation plan from all computing resource allocation plans; Adopting a reinforcement learning algorithm to iteratively optimize the resource allocation plan and adjusting the device-task allocation priority according to the feedback of historical task execution efficiency.

7. An adaptive method for dynamic task scheduling of distribution Internet of Things edge-cloud, according to claim 6, characterized in that The real-time monitoring of data transmission security and equipment status anomalies in the distribution network cloud computing environment and triggering a protection mechanism specifically includes: Implanting a two-way authentication protocol in the data transmission link to implement traffic rate limiting or blocking of abnormal access requests; Establish a device behavior baseline through an LSTM network. When the deviation of the operation sequence is detected to exceed the threshold, generate a risk warning and isolate the suspicious device; Regularly push security reports to the staff terminal.

8. An adaptive method for dynamic task scheduling of the distribution Internet of Things edge-cloud, according to claim 7, characterized in that The display of data processing results, task status, and security warning information through the gateway interface specifically includes: Render a 3D distribution network environment map in the gateway interface, and dynamically display the data streams of each device, the task execution progress, and the security status; Support voice and gesture interactions, and switch data views or adjust resource allocation plans according to the instructions of the staff; When a high-risk warning is triggered, automatically pop up an emergency handling interface and provide a multi-factor authentication channel.

9. An adaptive system for dynamic task scheduling of the distribution Internet of Things edge-cloud, characterized in that, An adaptive method for implementing the dynamic task scheduling of the distribution Internet of Things edge-cloud described in any one of claims 1-8, including: A data acquisition module, used to collect distribution network environment parameters and operation logs through temperature sensors, light sensors, and staff terminal devices, and add dynamic tags and access permission identifiers to sensitive data in the built-in encryption module; A preprocessing module, used to perform noise reduction, normalization, and real-time classification based on a convolutional neural network on the original data, and generate a structured data packet with a timestamp; A dynamic task allocation module, used to allocate computing tasks to the local GPU, cloud server, or other intelligent devices within the distribution network local area network through a load transfer algorithm according to the device load threshold and task type; A resource scheduling optimization module, used to iteratively generate the optimal allocation plan of network bandwidth and device priority based on the TOPSIS algorithm and reinforcement learning; A security protection module, used to implement abnormal traffic blocking and device isolation through a two-way authentication protocol and LSTM network behavior baseline detection; An interactive interface module, used to render a 3D distribution network environment map and support voice / gesture operations, and real-time display the task progress and the emergency verification interface for high-risk warnings.

Images