Tensor train based higher-order dynamic mode decomposition - A new big data mining algorithm in energy networks

Abstract

In today's increasingly demanding world of energy, the problem of big data analysis in intelligent energy networks (IENs) is becoming commonplace. Traditional methods of data analysis are mainly based on low-dimensional systems, which will become unfeasible in the future. Among the high-dimensional big data analysis methods that have been proposed, tensor-train decomposition (TTD) has become a research hotspot due to its inherent structural advantages. TTD makes it possible to analyse high-dimensional data independent of the "curse of dimensionality". There are already many applications that incorporate TTD and achieve good results. Due to the dynamic nature of IENs, we propose a new algorithm combining TTD and higher-order dynamic mode decomposition (HODMD), which is called tensor-train based higher-order dynamic mode decomposition (TT-HODMD), to demonstrate the superiority of HODMD for the analysis of dynamic systems and to extend it to situations where big data exist. The validation of the proposed TT-HODMD algorithm demonstrates the possibility of applying TT-HODMD to IENs. Finally, three real-life cases in IENs from Russia, China and the UK are prepared and the results of all three cases demonstrate the value and validity of TT-HODMD for use in IENs.

Date of Award	1 Aug 2023
Original language	English
Awarding Institution	The University of Manchester
Supervisor	Sergey Utyuzhnikov (Supervisor) & Chamil Abeykoon (Supervisor)