Recovering Missing Photovoltaic Signal (Gaps) With Low Dimensional Manifold Model
Abstract
An adaptation of the Low Dimensional Manifold Model--frequently used for image inpainting--is presented as an alternative for recovering signal (missing data) in photovoltaic generation measurements. This adaptation takes advantage of expected redundancies across simultaneous measurements to fill the gaps by assuming that these redundancies induce low local dimensionality in the mathematical structure (manifold) where multivariate observations are found. Experimental results with actual data from smart meters in Cape Verde corroborate the expected superiority of this approach, as compared to state of the art methods.
Keywords
Signal gaps, LDMM, Inpainting, Reverse Power