Load Profile Identification using Independent Component Analysis

Abstract

The paper aims to establish the load profiles for the consumers, in a looped network, when the power circuits are known on the branches The algorithm used is independent component analysis (ICA), a statistic computational method that allows the extraction of additive components from a mixed signal. The method requires as restriction the statistically independence of the additive components. In the paper, ICA was used to reconstruct some statistically independent signals (power requests according to consumer load profiles) from a mixture of signals (power flows on branches). The ICA algorithm does not require knowledge of network parameters or configuration. The consumer load profiles obtained by the independent component method overlapped with a very good approximation over the real load graphs of consumers used in modeling.