Power time series data retrieval method based on edge collaborative classification

Abstract

The invention belongs to the technical field of time series data retrieval, and particularly relates to an edge collaborative classification-based power time series data retrieval method, which comprises the following steps of: performing K-shape clustering on a power data set, and obtaining an optimal clustering number according to an elbow rule; creating a classification model by using a residual neural network model according to a clustering classification result; adopting federal learning to aggregate model parameters on different edge devices; when a user inputs time series data for retrieval, classifying the time series data by using the residual neural network model; dTW calculation is carried out on the data of each category, and the most approximate N pieces of time sequence data serve as retrieval results to be recommended to the user; according to the method, the problems of contradiction between resource requirements of an intelligent algorithm and limitation of edge devices to resources, contradiction between service quality and privacy protection and contradiction between various intelligent task requirements and single capability of the edge devices are solved, and collaborative classification and retrieval of different edge devices are realized.

Description

technical field [0001] The invention belongs to the technical field of time series data retrieval, in particular to a power time series data retrieval method based on edge cooperative classification. Background technique [0002] With the continuous advancement of applications such as smart grids, smart cities, and smart transportation, the number of connected devices and the data generated are exploding. The traditional cloud computing model, that is, centralized processing, has been unable to carry massive amounts of data, making the problems of cloud center transmission bandwidth, load, and data security more and more prominent. Computational models are born from applications. Time series data is one of the more common data types in the real world. It means a set of numbers that are sequentially observed and arranged on different time axes. It is widely used in meteorological data, industrial data, financial data, and power data. , medical data, sensor network monitorin...

AI Technical Summary

Problems solved by technology

However, there are still many contradictions in the process of migrating artificial intelligence algorithms from the cloud to the edge, such as the contradiction between the resource requirements of intelligent algorithms and the limited resources of edge devices, the contradiction between service quality and privacy protection, and the requirements of intelligent tasks. Contradictions Between Diversity and Single Capabilities of Edge Devices

Images