Estudo do controle ativo para aumentar a capacidade de hospedagem de sistemas fotovoltaicos em redes de distribuição do futuro

Author(s)

Guilherme de Oliveira Custódio

• Document type: Master's Dissertation
• Press: Campinas, SP.
• Institution: Universidade Estadual de Campinas (UNICAMP). Faculdade de Engenharia Elétrica e de Computação
• Defense date: 2021-03-23

Examining board members

Fernanda Caseño Trindade Arioli; Jonatas Boas Leite; João Paulo Papa

• Advisor: Fernanda Caseño Trindade Arioli

Abstract

Electrical power distribution networks are likely to experience in the next decade significant infrastructural changes due to the massive integration of low carbon technologies. The deployment of these devices, for instance, photovoltaic (PV) systems, are expected to be encouraged by governments in accordance with the Paris Agreement, signed by 195 countries, including Brazil, to reduce greenhouse emissions. For a certain penetration level, voltage transgressions, considering the statutory limits established by the Brazilian Electricity Regulatory Agency, and asset congestion will occur. Moreover, given a certain voltage level, PV systems will disconnect from the network, leading to loss of generation for customers. As a result, Distribution Network Operators (DNOs) management of future networks will be more challenging. In this context, such violations can be mitigated using typical controllable devices, for instance, on load tap changer transformers and line voltage regulators, along with other assets, such as PV system inverters. However, the adequate coordination of all these devices is difficult, requiring sophisticated control techniques. The main objective of this work is the investigation of the benefits and implementation challenges of a centralized active control based on AC Optimal power flow (OPF) Deep Reinforcement Learning that takes full advantage of typical and new controllable devices in future distribution networks to mitigate the impact of the massive penetration of PV systems and, thus, increase PV hosting capacity. The Active Control considers the unbalanced nature of distribution networks, since PV systems are mostly installed in low voltage networks. The techniques developed in this work will be tested using real distribution networks with installed PV systems and using machine learning applied to data of a Brazilian DNO (AU)

• FAPESP's process: 18/23391-9 - Active control to increase pv hosting capacity of electrical power distribution network
• Grantee: Guilherme de Oliveira Custodio
• Support Opportunities: Scholarships in Brazil - Master