A Consensus ADMM-Based Distributed Volt-VAr Optimization for Unbalanced Distribution Networks


Inaolaji A., Savasci A., Paudyal S., Kamalasadan S.

2022 IEEE Industry Applications Society Annual Meeting, IAS 2022, Michigan, Amerika Birleşik Devletleri, 9 - 14 Ekim 2022, cilt.2022-October identifier

  • Yayın Türü: Bildiri / Tam Metin Bildiri
  • Cilt numarası: 2022-October
  • Doi Numarası: 10.1109/ias54023.2022.9939862
  • Basıldığı Şehir: Michigan
  • Basıldığı Ülke: Amerika Birleşik Devletleri
  • Anahtar Kelimeler: alternating direction method of multipliers (ADMM), distributed algorithms, distribution systems, optimal power flow, Volt-VAr optimization
  • Abdullah Gül Üniversitesi Adresli: Hayır

Özet

Volt-VAr optimization (VVO) is usually performed by a central coordinator which provides the optimal setpoints of the control devices in the network for efficient voltage regulation in active distribution networks. However, such a central control scheme is prone to a single-point failure and results in privacy concerns. Conversely, distributed optimization methods decompose the entire network into subsystems such that local controllers compute a local optimization problem and have limited communication with their neighboring control agents, thereby enhancing data dignity. This work, therefore, adopts a distributed VVO approach which is based on the alternating direction method of multipliers (ADMM). The linearized Dist3Flow model (LinDist3Flow) is used as a convex grid model and the distributed VVO minimizes voltage deviation in an unbalanced active distribution system subject to limits on voltage magnitudes and inverter active and reactive power capabilities. Case studies implemented on the IEEE-123 node system and a 2522-node system demonstrate that ADMM-based VVO converges to the same solution as that obtained from the centralized VVO, but the performance of the ADMM algorithm is sensitive to the choice of the penalty parameter. Moreover, if the centralized formulation is convex and already tractable for even large-scale feeders, then solving the distributed counterpart might not necessarily provide any computational advantage but is still desirable for reasons such as data privacy and robustness to processor failure.