Smart grid is one of the major geo-distributed internets of things (IoT) networks. To support different functionalities of the smart grid such as continuous monitoring, the grid generates massive volumes of data. In a centralized system, majority of control decisions are accomplished at the cloud tier. This generates several drawbacks including limited bandwidth, privacy leakage, data confidentiality and integrity risk, and a single point of failure. Therefore, edge-computing paradigm is one of the possible solutions to avoid the drawbacks and make the system more trustworthy. This paper proposes an edge intelligence-based monitoring paradigm that can use data at the thing tier to monitor large variance shifts of control variables. We propose a Multivariate Exponentially Weighted Moving Variance (MEWMV) chart and a hybrid of wrapper and filter techniques to monitor the variables. The proposed approach can identify variables responsible for the out-of-control signals while considering the correlation among variables. This enables the grid to offload the decision task to the edge tier thus avoiding the latency, risks of data integrity and providing faster monitoring facilities. A case study on the smart grid indicates that the proposed hybrid can identify variables responsible for the out-of-control signals more accurately than existing approaches.
An edge tier task offloading to identify sources of variance shifts in smart grid using a hybrid of wrapper and filter approaches
Fortino G.
2021-01-01
Abstract
Smart grid is one of the major geo-distributed internets of things (IoT) networks. To support different functionalities of the smart grid such as continuous monitoring, the grid generates massive volumes of data. In a centralized system, majority of control decisions are accomplished at the cloud tier. This generates several drawbacks including limited bandwidth, privacy leakage, data confidentiality and integrity risk, and a single point of failure. Therefore, edge-computing paradigm is one of the possible solutions to avoid the drawbacks and make the system more trustworthy. This paper proposes an edge intelligence-based monitoring paradigm that can use data at the thing tier to monitor large variance shifts of control variables. We propose a Multivariate Exponentially Weighted Moving Variance (MEWMV) chart and a hybrid of wrapper and filter techniques to monitor the variables. The proposed approach can identify variables responsible for the out-of-control signals while considering the correlation among variables. This enables the grid to offload the decision task to the edge tier thus avoiding the latency, risks of data integrity and providing faster monitoring facilities. A case study on the smart grid indicates that the proposed hybrid can identify variables responsible for the out-of-control signals more accurately than existing approaches.I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.