Local reactive power dispatch optimisation minimising global objectives. (15th March 2020)

• Record Type: Journal Article
• Authors: Gandhi, Oktoviano; Zhang, Wenjie; Rodríguez-Gallegos, Carlos D.; Verbois, Hadrien; Sun, Hongbin; Reindl, Thomas; Srinivasan, Dipti
• Is Part Of: Applied energy. Volume 262(2020)
• Publication Date: 2020-03-15
• Subjects: Distribution system -- Local control -- Optimisation -- Photovoltaic systems -- Reactive power; Power (Mechanics) -- Periodicals; Energy conservation -- Periodicals; Energy conversion -- Periodicals; 621.042
• DOI: 10.1016/j.apenergy.2020.114529
• Languages: English
• ISSNs: 0306-2619

Abstract

Highlights: Local optimization of global variables is proposed. It does not require any communication with central controller or neighbors. The approach is stable and performs as well as centralized optimization. It does not need grid topology and removes privacy and latency issues. The importance of solar forecasting and algorithm validation is highlighted. Abstract: The number of distributed energy resources (DERs) deployed in distribution systems has been rapidly increasing in recent years. Consequently, many researchers have proposed to utilise the DERs for local reactive power support. Yet, the expansion of infrastructure necessary to allow communication among the DERs, and with a centralised controller, has not been as fast. This necessitates an independent local control method that is able to fulfil the objectives of a centralised controller. Therefore, this work proposes a data-driven local optimisation of global objectives (LOGO) algorithm to control the reactive power dispatch from DERs in distribution systems. The proposed method has been validated across different test systems with real topology, irradiance and load data, for steady-state undervoltage and overvoltage scenarios. The results have shown that LOGO performs almost as well as centralised optimisation without any communication and without any information regarding the grid topology. LOGO is able to satisfy the voltage constraints using only locally available information even when other algorithms fail to do so. Compared with distributed and other local reactive power controls, LOGO is much more stable and yields significantly better results. Moreover, LOGO's superiority over other control methods increases with system size and complexity. Hence, this work provides a viable alternative for real-life dispatch optimisation and opens up new possibilities to optimise global variables without any communication. Lastly, this paper also highlights the importance of accurate solar forecasting and validation of reactive power control.