SCADA-agnostic Power Modelling for Distributed Renewable Energy Sources

Abstract

<p>Distributed Renewable Energy Sources (DRES) are<br> considered as instrumental within modern smart grids and more<br> broadly to the various ancillary services contained within the<br> energy trading market. Thus, the adequate power production<br> profiling and forecasting of DRES deployments is of vital<br> importance such as to support various grid optimisation and<br> accounting processes. The variety of DRES installation companies<br> in conjunction with the diversity of ownership on DRES machinery,<br> controller firmware and Supervisory Control and Data<br> Acquisition (SCADA) software leads to cases where centralised<br> SCADA measurements are not entirely available or are provided<br> under a subscription-based model. In this work, we consider this<br> pragmatic scenario and introduce a SCADA-agnostic approach<br> that utilises freely available weather measurements for explicitly<br> profiling and forecasting power generation as produced in real<br> wind turbine deployments. For this purpose, we leverage various<br> machine learning (ML) libraries to demonstrate the applicability<br> of our system and further compare it with forecasting outputs obtained<br> when using SCADA measurements. Through this study, we<br> demonstrate a viable and exogenous profiling solution achieving<br> similar accuracy with SCADA-based schemes under much lower<br> computational costs.</p>