Research And Application Of Time-series Data Retrieval Method For Smart Grid

With the continuous improvement of power network intelligence and the increasing data sensing capability of power terminal devices,the scale of power timing data grows rapidly and plays an increasingly important role in the power network operation.Due to the characteristics of complexity,dynamism and large scale of power timing data,the retrieval of time series data has become a key challenge for research in this field.However,the timing data collected by power terminal devices often contain sensitive information related to users,and the processing of power timing data based on cloud computing method may leak users’ private information.Moreover,the existing time series data retrieval methods lack the consideration of the limited resources of power terminal devices,and cannot meet the demand of realistic grid application requirements.Therefore,how to protect user privacy while data retrievals has become the main challenge of current research.To address the current demand of privacy-preserving terminal power timing data retrieval and considering the multivariate nature of power timing data,this thesis investigates the privacy-preserving edge timing data retrieval problem from both ciphertext and plaintext aspects.In terms of cipher-text power timing data retrieval,a multi-key cipher-text retrieval method for intelligent terminal edge computing is proposed.In the plaintext power timing data retrieval,method framework of time series data retrieval based on edge collaborative classification is proposed.The following two main research contents are included.1.Against traditional searchable encryption schemes only focus on text-based data and lack of research on temporal data,we firstly propose the idea of effectiveness of multi-keyword cipher-text retrieval methods for smart grid edge computing,and design a research framework based on pseudo-random functions to provide basic support for privacy protection of edge power temporal data.Secondly,the fuzzy keyword index structure based on N-gram and Sim-Hash model is proposed to realize the fuzzy multikeyword cipher-text retrieval of power time series data.Finally,the performance of multikeyword cipher-text retrieval of edge power timing data is evaluated by index construction time,trapdoor generation time and user retrieval time.The experimental results show that the multi-keyword cipher-text retrieval scheme proposed in this thesis improves the retrieval efficiency of users while ensuring the accuracy of retrieval results.2.In view of the lack of consideration of edge device diversity and its limited resources in the current timing data retrieval scheme,a federated learning-based cooperative power timing data retrieval framework for sensitive data is proposed from the perspective of edge terminals,which performs edge power timing data classification modeling by K-shape clustering algorithm and neural network,and designs a federated learning-based cooperative power grid terminal timing data algorithm to meet the requirements of different mechanisms.Based on this,the lower bound algorithm is used to compare the similarity between the categories of input data,and data of the same category on different edge nodes,so as to realize the time series data retrieval of edge collaboration.The experimental results show that the efficiency of users’ similarity retrieval of power timing data using the lower bound algorithm is higher than that of the dynamic time regularization algorithm,which proves that the terminal power timing data retrieval scheme proposed in this thesis achieves efficient retrieval while ensuring the users’ privacy.This thesis validates the proposed scheme and algorithm using real power timing dataset.The experimental results show the effectiveness of the multi-keyword cipher-text retrieval scheme for smart grid edge computing and verify the feasibility of the temporal data retrieval scheme based on edge collaborative classification.