Electrical Model-Free Voltage Calculations Using Neural Networks and Smart Meter Data

The proliferation of residential technologies such as photovoltaic (PV) systems and electric vehicles can cause voltage issues in low voltage (LV) networks. During operation, voltage calculations can help determining settings, such as PV curtailment, that ensure compliance with statutory limits. In planning, voltage calculations can help assessing the effects of new connection requests. However, the main challenge for distribution companies is that voltage calculations require power flow analyses (or similar) that need accurate electrical models which, for LV networks, are not readily available for most companies. This paper proposes a scalable electrical model-free voltage calculation methodology that uses Neural Networks to capture the underlying relationships among historical smart meter data (
$\boldsymbol {P}$
,
$\boldsymbol {Q}$
, and
$\boldsymbol {V}$
) and the corresponding LV network. The methodology is demonstrated using half-hourly data from one realistic LV network (146 customers) from Victoria, Australia, over three weeks and with 20% PV penetration. To account for upstream voltage fluctuations, an integrated MV-LV network with 3,400+ customers is considered. Results considering different weeks with the same and higher PV penetrations demonstrate an average error of less than 2 Volts; showing that, in the absence of LV electrical models, the methodology could be used by distribution companies for different applications.