This paper deals with voltage regulation in medium voltage distribution networks with high penetration of distributed photovoltaic generation. More specifically, it considers electronic on-load tap changers and line voltage regulators as voltage control mechanisms operated exclusively by a machine learning algorithm. The considered network is a real-life 10 kV feeder with 41 equivalent loads and nine equivalent photovoltaic generators. Two years of real smart meters hourly data are considered for the feeder's loads and, together with simulated power generation data of photovoltaic systems, are used as inputs to a deep neural network. The data are first fed to a PowerFactory model of the feeder, and then, using quasi-dynamic analysis, the correct tap positions of electronic on-load tap changers and line voltage regulators are used as outputs. These inputs and outputs are finally utilized for training the deep neural network, which is expected will predict the outputs well when fed by new inputs never seen before.

Voltage regulation in distribution networks in the presence of distributed generation: LVR and E-OLTC with a machine learning approach

Bosisio A.
;
2022-01-01

Abstract

This paper deals with voltage regulation in medium voltage distribution networks with high penetration of distributed photovoltaic generation. More specifically, it considers electronic on-load tap changers and line voltage regulators as voltage control mechanisms operated exclusively by a machine learning algorithm. The considered network is a real-life 10 kV feeder with 41 equivalent loads and nine equivalent photovoltaic generators. Two years of real smart meters hourly data are considered for the feeder's loads and, together with simulated power generation data of photovoltaic systems, are used as inputs to a deep neural network. The data are first fed to a PowerFactory model of the feeder, and then, using quasi-dynamic analysis, the correct tap positions of electronic on-load tap changers and line voltage regulators are used as outputs. These inputs and outputs are finally utilized for training the deep neural network, which is expected will predict the outputs well when fed by new inputs never seen before.
2022
978-1-6654-7146-6
File in questo prodotto:
Non ci sono file associati a questo prodotto.

I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.

Utilizza questo identificativo per citare o creare un link a questo documento: https://hdl.handle.net/11571/1469975
Citazioni
  • ???jsp.display-item.citation.pmc??? ND
  • Scopus 0
  • ???jsp.display-item.citation.isi??? ND
social impact