Transformer-Based PV Capacity Estimation for Reactive Power Optimization in Active Distribution Networks

Abstract

To address inaccuracies in reactive power optimization caused by the absence of integrated distributed photovoltaic (PV) in active distribution network (ADN), this paper proposes a Transformer-based PV capacity estimation and reactive power optimization method for ADN with limited information. The self-attention mechanism is employed to capture long-term mappings between net-load variations and meteorological conditions. The supervised capacity loss and netload reconstruction loss variables are both introduced to improve the estimation accuracy of the PV capacity. The unit-capacity PV output potential is also involved to improve the interpretability of the model. A reactive power optimization model coordinating capacitor banks (CBs) and known-capacity PV inverters is established based on the estimated capacities of the PV nodes to improve the voltage profile. Case studies demonstrate that the proposed method provides effective data support for ADN reactive power optimization without PV capacity information.