| With the integration application of IoT technology and power distribution network,the development trend is forming to build a cloudedge-end collabortive distribution IoT system for improving the viewable,measurable,and controllable capability of power distribution network.The heterogeneous computing resource collaboration and IoT data fusion have significant research value,which can effectively improve the business processing reliability and accuracy,and enhance the intelligence level of power distribution network.Scholars and research institutions currently refer the urban IoT and Internet technologies to carry out industry-customized research about heterogeneous computing resource collaboration and IoT data fusion of business scenarios,such as intelligent distribution station area,smart power distribution room,distributed new energy grid connection.However,computing resource collaboration and IoT data fusion still face problems such as non-uniform basic information model,low resource collaboration efficiency,and poor data fusion quality as the scale of power distribution network expands,including the widely deployment of intelligent power distribution terminals and sensors,exponentially growing of data number,increasingly abundance of data type.Facing the reliability and accuracy of power distribution business processing,this manuscript conducts in-depth research on the computing resource collaboration and IoT data fusion method of power distribution IoT from four aspects:information model architecture design,heterogeneous resource allocation optimization,business bearing reliability guarantee,and IoT data fusion governance.The main research contents and innovations are as follows.(1)In view of the lack of resource collaboration and data fusion model due to the unified information model of power distribution IOT,a "sixdomain" power distribution IoT information architecture with equivalent,nestable and trimmable features is constructed.The information model and interaction logic between entities in the architecture are designed,the creating method of terminal information model is proposed,and the relevant industrial and national standards are published.Furthermore,the typical application scenarios and business processing flow of power distribution IoT system are designed following the standards,which provides the information model for computing resource collaboration and IoT data fusion.(2)To address the inefficient collaboration problem of multidisciplinary edge heterogeneous computing resources,a cloud-edge scheduling mechanism of power distribution IoT resources based on the differential game is proposed,considering the business priority,computing power cost and network quality with industrial control attributes.A cloudedge computing resource scheduling model is constructed to analyze the Nash equilibrium of computing resource scheduling feedback,which balance the asymmetric relationship between solidified computing resources and distributed power distribution IoT business,and optimize the allocation of power distribution IoT cloud-edge computing resources.Compare to stackelberg game model algorithm,the simulation results show that the proposed mechanism reduced the system overhead by 9%-15%for the edge IoT agent,and the experimental results show that the proposed mechanism reduced the system over-head by 3.3%for the edge IoT agent,which effectively improves the collaborative efficiency of the edge heterogeneous solidified computing resources.(3)Aiming at unbalanced load and decreased reliability problem due to the high concurrency of power distribution business under limited resource conditions,an edge-edge computing resource optimization method is proposed based on an improved ant colony algorithm.Combinatorial optimization model of dynamic unloading of edge-edge collaborative tasks is constructed to minimize the execution delay of task transmission and computational consumption,constrained by the service quality evaluation indexes such as CPU utilization,power consumption,memory utilization,external storage utilization,and node security level.Furthermore,a task allocation method is proposed,which could timely respond to high concurrency or burst services of power distribution business,and compress the running time window to avoid the reliability decrease caused by the overload of single node resources.Simulation results show that the proposed method has better performance in task execution time and task average waiting time,and CPU and memory average utilization reduce to less than 30%,compared to the modified round robin algorithm(RRA)and generalized priority algorithm(GPA).Experiment results show that the proposed method reduces the CPU utilization by 12%on average compared to RRA and GPA,and effectively improves the business carrying reliability of edge IoT agent.(4)For the poor data quality problems of power distribution collection caused by low accuracy of business processing,an IoT data fusion method for different edge computing nodes is proposed based on the collaboration of power distribution IoT computing resource.The nearest neighbor sample points to be classified are given higher weights,and a data-missing repair model based on clustering and correlation analysis is established.Furthermore,a edge-edge data fusion method oriented to small sample learning is designed and applied to the fault diagnosis of power distribution metering,which improves the quality of power distribution data and the accuracy of business processes.Experiment results show that the smaller clustering K value is,the better corresponding repair effect.With 1%proportion of missing data,the repair effect is improved by more than 81%compared to the KNN algorithm.With 5%proportion of missing data,the repair effect is improved by more than 44.7%.Meanwhile,the proposed method has an accuracy rate of 85.2%under the condition of fewer samples,which increases by 40.6%compared to CNN methods. |
