Data-driven combined central and distributed volt/var control in active distribution networks

Abstract

This paper proposes a two-stage combined central and distributed Volt/Var control (VVC) strategy to coordinate both utility-scale and customer-owned photovoltaics (PVs) for voltage regulation in active distribution networks (ADN). In the first stage, with day-ahead PV and load predictions, the central controller (CC) performs optimal power flow (OPF) to determine the hourly dispatch of on-load tap changers (OLTC) and capacitor banks (CBs). In the second stage, the ADN is partitioned into different zones only based on PV and load types. The zone controllers (ZC), which is formulated by deep neural networks (DNN), then communicate with their neighbors and track the optimal total reactive power support of PVs in each zone. In the offline training process, an analytical derivation to minimize power losses is also provided that can help the input and output selection of DNN. To online dispatch reactive power for individual utility-scale and customer-owned PVs in corresponding zones, a multi-mode reactive power allocation strategy is proposed. Three different modes namely loss reduction mode (LRM), fair power sharing mode (FPM) and voltage security mode (VSM) are designed to meet varying operation requirements, which can mitigate voltage violations with fair participation for customers and reduced power losses for system operators. Numerical simulations are performed on the IEEE 33-bus distribution system to demonstrate the effectiveness of the proposed method in finding a trade-of…