Blind Source Separation in Sustainable Energy Systems Using Free Component Analysis Considering Power-usage Interdependence

Abstract

Blind source separation is crucial for improving situational awareness in modern energy systems, particularly a sustainable one with extensive integration of distributed energy resources (DERs). DER power-usage patterns, shaped by climate and social factors, exhibit significant interdependence. This poses challenges for traditional methods such as independent component analysis, which rely on assumptions of independence. In this context, our work proposes a free component analysis (FCA) framework. FCA, rooted in free probability and random matrix theory, employs (free) non-commutative matrix variables, departing from traditional (independent) scalar variables. This approach effectively captures spatial-temporal correlations, offering deeper insights into DER cluster behaviours and further informing decision-making within sustainable energy systems featuring coupled DERs. Case studies using both simulated data and field data validate the effectiveness of the proposed framework.