A Data-Driven Methodology for Modeling Losses in HTS Power Systems

João Murta-Pina, Roberto André Henrique De Oliveira, Anabela Gonçalves Pronto, Henrique Simas, Isabel Catarino, João Rosas, Masoud Ardestani, Alfredo Álvarez, Pilar Suarez, Belén Rivera

Research output: Contribution to conferencePosterpeer-review

Abstract

Calculation of losses in the design of HTS devices requires the use of complex, computational intensive numerical modelling tools, applied in different formulations/state variables, with different accuracy and requirements Typically, just a section of the system is represented, often reducing dimensionality, to decrease the required computations This is due to nonlinear resistivity of HTS (current field temperature dependent), and extreme aspect ratios, where characteristic dimensions range from micrometers of HTS tapes’ thickness to hundreds of meters of, e g cables length. Those numerical tools thus require huge computation times which make their application often unfeasible in the design of large scale devices (as changing parameters, e g number of turns in windings, would require running whole new simulations) or in the integration into grid simulation software (to assess the performance of the devices in realistic operating conditions and their interaction with other grid elements). Besides these limitations, the calculation of losses often only addresses pure AC regimes, but grid currents are harmonically polluted. This work proposes a hierarchical, data driven approach for modeling losses in HTS power systems The problem is split into several levels, from HTS tapes to the whole device, where higher levels receive inputs from lower ones By mixing empiric, equation based models with black box modelling, namely artificial neural networks, the computation times required to the preliminary evaluation of losses is reduced unprecedentedly.
Original languageEnglish
Publication statusPublished - 9 Sep 2021
Event15th European Conference on Applied Superconductivity - Moscow, Russian Federation
Duration: 5 Sep 20209 Sep 2021
Conference number: 15
https://www.eucas2021.org

Conference

Conference15th European Conference on Applied Superconductivity
Abbreviated titleEUCAS
Country/TerritoryRussian Federation
CityMoscow
Period5/09/209/09/21
Internet address

Fingerprint

Dive into the research topics of 'A Data-Driven Methodology for Modeling Losses in HTS Power Systems'. Together they form a unique fingerprint.

Cite this