Estimating PV Curtailed Power as a Voltage Support Service using Data-Driven Approaches

Conference paper (2022)

Authors

Gijs Verhoeven Eindhoven University of Technology
P.P. Vergara Barrios Intelligent Electrical Power Grids - EEMCS
Edgar Mauricio Salazar Eindhoven University of Technology
Koen Kok Eindhoven University of Technology
Research Group
Intelligent Electrical Power Grids (EEMCS) (TU Delft)
Copyright: © 2022 Gijs Verhoeven, P.P. Vergara Barrios, Edgar Mauricio Salazar Duque, Koen Kok
DOI
help_outline
https://doi.org/10.1109/SEST53650.2022.9898486
Machine learning Droop control Overvoltage Low voltage distribution systems PV curtailment
More Info
expand_more

Published Date

2022

Language

English

Reuse Rights

Other than for strictly personal use, it is not permitted to download, forward or distribute the text or part of it, without the consent of the author(s) and/or copyright holder(s), unless the work is under an open content license such as Creative Commons.

Faculty

Electrical Engineering, Mathematics and Computer Science

Department

Electrical Sustainable Energy

Research Group

Intelligent Electrical Power Grids

Abstract

To guarantee a successful deployment of a droop-based control strategy to mitigate overvoltage problems caused by solar photovoltaic (PV) generation, Distribution System Operators (DSOs) will need to estimate the amount of active power curtailed by the PV inverters for billing purposes. This paper provides a structural elaboration on the development of data-driven approaches in Python to estimate the PV curtailed power as a provision of voltage support services by residential users using droop-based voltage control strategies. The use of the total input data, available for a DSO, would be impractical for an all-regression approach for the estimation of the PV curtailed power. Since in the majority of the data no active power is curtailed, the data-driven models would in this case partly be trained and fitted for situations where there is no active power curtailment. The regression models for the curtailed power prediction are therefore preceded by a classification model. The developed combined classification-regression model was able to estimate the PV curtailed power with an error of less than 4%, for test data from the network on which the model was trained.