Cluster-Based Volt/Var Optimization on a Utility Distribution Feeder with Forecasted EV Penetration

With significant increase in EV adoption expected in the near future and the associated impacts on power systems, the effect of volt/var optimization (VVO) as the next step to conservation voltage reduction (CVR), requires reevaluation. Implementation of a cluster-based VVO control strategy employs a novel approach with machine learning (ML) load forecasting to reduce device adjustments through <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\mathrm{k}$</tex>-means clustering of contiguous time steps with similar active power load in which a single adjustment would be sufficient. The cluster-based VVO method is tested on a complex real world utility distribution feeder with 2,018 nodes, 8.65MW peak load, 9 capacitor banks (CBs), and a load tap changer (LTC) at the substation transformer. Performance of the cluster-based VVO method for the circuit with high forecasted EV penetration is tested with comparison to a baseline case. Total reduction in energy consumption of 1.7% and 1.9% in the expected range of 1.4% with minimal tap changes over 24 hours was achieved by the cluster-based VVO method with and without EV charging, respectfully.