Load Library Construction and Self-learning Method Based on Cloud-Edge Collaboration

Abstract

With the access of various loads and distributed sources, the load composition and features are complicated. In order to grasp the load composition of the distribution area and carry out advanced applications, the load library is established. However, the poor generalization ability of the load library leads to difficulty adapting to different distribution networks. This paper proposes a load library construction and self-learning method based on Cloud-Edge collaboration. Firstly, a Cloud-Edge collaboration is proposed by distributing the data and resources to improve the generalization ability of the load library. Secondly, a cloud-side load library construction method is proposed by analyzing the representative load features. A data-driven load identification is carried out based on load features and cloud-side load library. Then, a self-learning method based on the deviation degree is proposed. The deviation degree is calculated online in the edge to correct the cloud-side load library. Finally, the proposed methods are verified the feasibility using actual load data. The results show that the scheme proposed in this paper improves the adaptability of cloud-side load library in different distribution networks, and improves the accuracy of advanced applications based on cloud-side load library.