A4 Vertaisarvioitu artikkeli konferenssijulkaisussa
Context Aware Monitoring for Smart Grids
Tekijät: Hauer Daniel, Götzinger Maximilian, Jantsch Axel, Kintzler Florian
Toimittaja: N/A
Konferenssin vakiintunut nimi: IEEE International Symposium on Industrial Electronics
Julkaisuvuosi: 2021
Journal: Proceedings of the IEEE International Symposium on Industrial Electronics
Kokoomateoksen nimi: 2021 IEEE 30th International Symposium on Industrial Electronics (ISIE)
Tietokannassa oleva lehden nimi: PROCEEDINGS OF 2021 IEEE 30TH INTERNATIONAL SYMPOSIUM ON INDUSTRIAL ELECTRONICS (ISIE)
Lehden akronyymi: PROC IEEE INT SYMP
Sarjan nimi: Proceedings of the IEEE International Symposium on Industrial Electronics
Sivujen määrä: 6
ISBN: 978-1-7281-9024-2
eISBN: 978-1-7281-9023-5
ISSN: 2163-5137
eISSN: 2163-5145
DOI: https://doi.org/10.1109/ISIE45552.2021.9576488
Verkko-osoite: https://ieeexplore.ieee.org/document/9576488
Today's energy grids face an increasing number of decentralized and renewable energy sources as well as growing e-mobility. Therefore, reliable grid monitoring becomes a key element for a sustainable grid operation. Traditional grid monitoring concepts are either fully manual, need a detailed system model, or rely on computationally heavy machine learning concepts. However, with the given complexity of the energy grid, a model-free and context-aware monitoring approach can save resources and efforts. Recently, we introduced the Confidence-based Context-Aware Condition Monitoring (CCAM) system and successfully tested it on two different industrial use-cases: a hydraulic circuit and an AC motor. In this paper, we enhance CCAM for a third, entirely different industrial use case, an energy grid, by introducing two extensions - a continuous reevaluation and a state mooring approach. Furthermore, we present a new Smart Grid monitoring methodology on top of CCAM, paving the way for new real-time grid control systems. We evaluate our approach based on historical load data from a low voltage grid section. Our results show that characteristics of a daily load profile can be learned and outliers can be detected.
